Patent Application
20210406735
This invention relates to a document processing and response generation system, and in particular to an approach to facilitate quick and efficient determination of answers from a potentially large repository of documents in response to user queries, which may be provided as natural language questions.
Organizations often collect large number of documents to service as a repository of information, be it administrative of technical information. For example, a corporation may have a large library of human resource documents, which together define, in a hopefully consistent manner, the HR policies and procedures of the corporation. A user, such as a corporate employee, may have to search the collection of documents to answer a question, such as “how much vacation time am I entitled to?”
A common problem with file-centric processing where unstructured and structured data exist in the file is that a rich natural language interface does not exist for users to navigate the document according to their needs and desires. As such, for example, many documents have long pages of explanation that are in small font that, if needed, force the user to read the entire explanation to find the key information the user needs.
Another problem that may occur in situations involving a large library of documents is that the organization may not have the resources to arrange and manage their documents into easily searchable documents (i.e., for question-answer querying), and may need to rely on a third-party's services to do so. In yet another situation, organizations that have the resources to maintain and manage their documents in searchable forms, may have multiple distributed computing networks (at different locales), and may decide (for efficiency and cost reasons) to store searchable documents at a single centralized location instead of individually storing the same set of searchable documents at each of their separate local networks. In such situations, data security and privacy issues become a concern, especially if sensitive information needs to be communicated across communication media accessible to people outside the organizations storing their documents at a central, or 3rd-party, searchable document repository.
The present disclosure is directed to a document processing and response generation system configured to determine an answer, from a library of documents and various other data management applications, accessible by the document processing system, in response to a query provided by a user. In some embodiments, the system returns, to the user that submitted the query, output that includes a pointer to a part of a document (e.g., a phrase or an excerpt) in the library of documents (or in stored records of the data management applications), that is separately accessible by the user, so that the user can retrieve the document (and/or specific portions of the document) pointed to by the pointer. The actual user-readable answer and/or a summary of the answer may additionally or alternatively be provided to the user. The document library can be a remote secure data repository that the document processing system is credentialed to access, or can be a local repository. Document processing may be based on one or more transformations of source content into transformed content that is arranged in a document object model (DOM) item. Q-A searching can then be applied to the transformed content. Search accuracy and speed can be improved through use of a question-answer cache.
In some variations, a method is provided that includes receiving, at a local device from a remote device, query data representative of a question relating to source content of a source document, and determining whether one or more pre-determined questions stored in a question-answer cache maintained at the local device matches the query data according to one or more matching criteria. The method further includes obtaining from the question-answer cache, in response to a determination that at least one of the pre-determined questions matches the query data received from the remote device, at least one answer data item, associated with at least one pre-determined question, corresponding to an answer to the question relating to the source content.
Embodiments of the method may include at least some of the features described in the present disclosure, including one or more of the following features.
Obtaining from the question-answer cache at least one answer data item may include obtaining at least a pointer to access, from a copy of the source document accessible from the remote device, at least one portion of the copy of the source document corresponding to the answer to the question relating to the source content.
Obtaining from the question-answer cache at least one answer data item may include obtaining one or more of, for example, at least part of the source content, and/or a summary of the at least one portion of the source content.
The method may further include storing in the question-answer cache a data item representative of a pair of a particular question and a corresponding answer in response to receiving an indication, from a user, that a portion of the data item representative of the corresponding answer represents a satisfactory response to the particular question.
The method may further include obtaining, at the local device, replacement answer data to replace an initial answer data determined at the local device in response to a particular question, with the initial answer being determined by a user to be an unsatisfactory response to the particular question, and storing in the question-answer cache a data item representative of a pairing of the particular question and the replacement answer.
The method may further include searching at the local device, in response to a determination that none of the pre-determined questions stored in the question-answer cache matches the query data received from the remote device, a documents storage, comprising a plurality of document objects of transformed content corresponding to one or more source documents, to identify one or more of the plurality of document objects according to the query data, and generating, in response to determination, at the local device, of a match between the query data and at least one portion of the transformed content, output data based on the at least one portion of the transformed content.
The output data may include at least a pointer to access, in a copy of the source document accessible from the remote device, at least one portion of the copy of the source document corresponding to the at least one portion of the transformed content.
The method may further include storing the plurality of document objects at the document storage, including receiving one or more source documents at the local device, segmenting the one or more source documents into a plurality of document segments, transforming each of the plurality of document segments into one or more respective transformed segments according to one or more transformations, and arranging the transformed segments into the plurality of document objects.
The method may further include identifying, based on the received query data representative of the question, one or more additional questions from a repository of query data representative of a plurality of questions associated with source documents, and generating additional output data comprising additional answer data responsive to at least one of the identified one or more additional questions.
The method may further include obtaining the one or more additional questions based on one or more of, for example, input data received from one or more users of the local device and/or the remote device, and/or metadata associated with identified responses to the query data determined from document objects, stored at the local device, associated with the source document. In such embodiments, the obtaining may also include storing in the repository of query data one or more additional question data items representative of the one or more additional questions.
Determining whether the one or more pre-determined questions stored in the question-answer cache maintained at the local device matches the query data according to one or more matching criteria may include determining linguistic meaning of the query data, determining a level of meaning similarity between the query data and respective meaning of at least some of the pre-determined questions in the question-answer cache, and determining that at least one of the one or more pre-determined questions is a match to the query data when the level of meaning similarity between the query data and the respective meaning for the at least one of the one or more pre-determined questions exceeds a pre-determined similarity threshold.
The method may further include determining that the query data matches one of pre-determined commands associated with respective processes performed, at the local device, with respect to the source document, and executing the pre-determined process associated with the one of the pre-determined commands determined to match the query data.
Determining that the query data matches one of pre-determined commands may include determining that the query data matches a summary command configured to cause generation of a summary of at least a portion of the source document.
The method may further include generating one or more augmented questions based on the question corresponding to the query data.
Generating the one or more augmented questions may include revising the question represented by the query data based on one or more of, for example, a) changing order of wording in the question, and/or b) substituting one or more words in the question with substitute words identified from a synonym data record to be synonyms or equivalent words to the one or more words in the question.
The method may further include searching, at the local device, a documents storage, comprising a plurality of document objects of transformed content corresponding to one or more source documents, to identify one or more of the plurality of document objects according to the one or more augmented questions.
Determining whether the one or more pre-determined questions stored in the question-answer cache maintained at the local device matches the query data according to one or more matching criteria may include transforming the query data into a transformed query according to one or more transformations, and matching the transformed query to transformed Q-A items stored in the question-answer cache. Each of the transformed Q-A items may include transformed data resulting from transformation of one or more questions and corresponding one or more answers according to the one or more transformations.
The transformed Q-A items may include the transformed data resulting from a Bidirectional Encoder Representation from Transformers (BERT)-based transformation of the one or more questions and the corresponding one or more answers.
In some variations, a system is provided that includes a communication unit configured to receive, from a remote device, query data representative of a question relating to source content of a source document, a question-and-answer cache storage device configured to store pre-determined questions, and a controller (which may be the same or part of the query processing module described below in relation to
Embodiments of the system may include at least some of the features described in the present disclosure, including at least some of the features described above in relation to the method, as well as one or more of the following features.
The controller configured to obtain from the question-answer cache at least one answer data item may be configured to obtain one or more of, for example, a pointer to access, from a copy of the source document accessible from the remote device, at least one portion of the copy of the source document corresponding to the answer to the question relating to the source content, at least part of the source content, and/or a summary of the at least one portion of the source content.
The controller may further be configured to cause storage, in the question-answer cache, of a data item representative of a pair of a particular question and a corresponding answer in response to receiving an indication, from a user, that a portion of the data item representative of the corresponding answer represents a satisfactory response to the particular question.
The controller may further be configured to obtain, at the local device, replacement answer data to replace an initial answer data determined at the local device in response to a particular question, with the initial answer being determined by a user to be an unsatisfactory response to the particular question, and cause storage in the question-answer cache of a data item representative of a pairing of the particular question and the replacement answer.
The controller may further be configured to search, in response to a determination that none of the pre-determined questions stored in the question-answer cache matches the query data received from the remote device comprising a plurality of document objects of transformed content corresponding to one or more source documents, to identify one or more of the plurality of document objects according to the query data, and generate, in response to determination of a match between the query data and at least one portion of the transformed content, output data based on the at least one portion of the transformed content.
The controller may further be configured to identify, based on the received query data representative of the question, one or more additional questions from a repository of query data representative of a plurality of questions associated with source documents, and generate additional output data comprising additional answer data responsive to at least one of the identified one or more additional questions.
The controller configured to determine whether the one or more pre-determined questions stored in the question-answer cache matches the query data according to one or more matching criteria may be configured to determine linguistic meaning of the query data, determine a level of meaning similarity between the query data and respective meaning of at least some of the pre-determined questions in the question-answer cache, and determine that at least one of the one or more pre-determined questions is a match to the query data when the level of meaning similarity between the query data and the respective meaning for the at least one of the one or more pre-determined questions exceeds a pre-determined similarity threshold.
The controller may further be configured to determine that the query data matches one of pre-determined commands associated with respective processes performed with respect to the source document, and executing the pre-determined process associated with the one of the pre-determined commands determined to match the query data.
The controller configured to determine that the query data matches one of pre-determined commands may be configured to determine that the query data matches a summary command configured to cause generation of a summary of at least a portion of the source document.
The controller may further be configured to generate one or more augmented questions based on the question corresponding to the query data, including to revise the question represented by the query data based on one or more of, for example, a) changing order of wording in the question, and/or b) substituting one or more words in the question with substitute words identified from a synonym data record to be synonyms or equivalent words to the one or more words in the question.
The controller configured to determine whether the one or more pre-determined questions stored in the question-answer cache matches the query data according to one or more matching criteria may be configured to transform the query data into a transformed query according to one or more transformations, and match the transformed query to transformed Q-A items stored in the question-answer cache, wherein each of the transformed Q-A items includes transformed data resulting from transformation of one or more questions and corresponding one or more answers according to the one or more transformations.
In some variations, a non-transitory computer readable media is provided, that is programmed with instructions, executable on one or more processors of a computing system, to receive, at a local device from a remote device, query data representative of a question relating to source content of a source document, determine whether one or more pre-determined questions stored in a question-answer cache maintained at the local device matches the query data according to one or more matching criteria, and obtain from the question-answer cache, in response to a determination that at least one of the pre-determined questions matches the query data received from the remote device, at least one answer data item, associated with at least one pre-determined question, corresponding to an answer to the question relating to the source content.
Embodiments of the computer readable media may include at least some of the features described in the present disclosure, including at least some of the features described above in relation to the method and to the system.
Other features and advantages of the invention are apparent from the following description, and from the claims.
These and other aspects will now be described in detail with reference to the following drawings.
Like reference symbols in the various drawings indicate like elements.
With reference to
In some embodiments, searching the document object repository to find an answer to a query typically includes two operations: (1) first, a process referred to as Fast-Search or Fast Match (FM) process is performed, and (2) the Fast-Match process is then followed by a process called Detailed-Search or Detailed-Match (DM) process (also referred to herein as “fine-detail” search). Both the FM and DM processes can be based on BERT (Bidirectional Encoder Representations from Transformers) models. In the FM case, the model results (in some implementations) in, for example, one vector for a query and one vector for one paragraph (e.g., 200 words window, which may also include contextual data). In the DM, there are typically multiple vectors per query or per paragraph, in proportion to the number of, for example, words or sub-words, in the query or paragraph.
It is to be noted that, in some embodiments, the transformations of the query and/or the source documents may be performed at a customer's network, with the transformed query and/or transformed content then communicated to a central server. Such embodiments can improve privacy and security for communicating sensitive data across networks since resultant vectors (derived through the transformation of content or query data) are created in the secure space of the customer (client), and consequently only the resultant transformed vectors (rather than the actual content or query data) are available or present at the centralized cloud server. The transformation of the content or query data at the client's device can act as a type of encryption applied to the data being transformed and will thus result in secure processing that protects the data from attacks on the server cloud. In some embodiments, the data being transformed at the client's network can additionally be encrypted to provide even further enhanced secured communication of the client's data (be it source data or query data).
The general architecture of a document processing and searching system 100 and the processes it implements is next provided. Intricacies of the implementations details for the various modules, units, and processes of
Yet in other alternative embodiments, some portions of the system (e.g., the ingestion units configured to perform the pre-processing and vectorization operations on source documents and/or on queries submitted by users) may be located inside the firewall of a customer's network, while storage of ingested document (and optionally search engines to search ingested content) may be located outside the customer's network's firewall (e.g., on a centralized cloud server(s)). In such alternative embodiments, data sent to the cloud servers (e.g., to perform the search at a centralized location) may already have been processed into encoded (ingested) content (e.g., through vector processing that may have been implemented through coarse transform, e.g., applied to fixed sized input segments, and/or fine-detail numerical transforms applied to smaller portions than the portions processed by the coarse transformer) that is unintelligible to third parties unauthorized to make use of the data, thus adding another measure of privacy and security protection to data that is to be processed using the system 100. In these alternative embodiments, the initial part of the processing of the input query may also be processed inside the customer network's firewall. In addition to performing the transformation (of the source content and/or the query) within a client's firewall, such transformed data may further be encrypted (using symmetric or asymmetric encryption keys) before being transmitted to the document processing agent 110, thus increasing the level of security/privacy realized for communications between a customer's network and the centralized document processing agent (which serves multiple customers).
In some embodiments, the example customer network 150a may be a distributed set of stations, potentially with a dedicated secured gateway (protected by a firewall and/or other security measures) that can be controlled (from a station 152) by an administrator. The customer generally has amassed a large volume of electronic documents (including, e.g., technical documentation relevant to the customer's operations, administrative documents such as Human Resource documents, and all other types of written documents in electronic form). The documents are arranged in a document library 160 (which may be part of the computing of the customer network 150a), and are accessible by various authorized users at user stations 154a-c within the network 150a, and by administrator (via an administrator station 154). Any number of stations may be deployed in any particular customer network/system. The administrator station 152 can control access to the documents in the library 160 by controlling privileges, and otherwise managing the documents (e.g., access to specific documents within the library 160, management of content to conceal portions that do not comply with privacy requirements, etc.) As will be discussed in greater detail below, in addition to the library 160 (containing documents relating to operation of the entity operating on the network), other sources of data or information may be available from various applications employed by the customer (e.g., an e-mail application, a chat application such as Slack, customer relationship applications such as Salesforce, etc.) to process through the document processing implementations described herein.
The administrator station 152 is configured to communicate with the document processing agent 110 via, for example, an admin interface 125. Among other functions, the administrator can provide the document processing agent 110 with information identifying location of the source document in the repository (library) 160 maintaining the plurality of source documents, control configuration and operation of the functionality of the document processing agent 110 in relation to the customer network 150a, review data produced by the agent 110 (e.g., override certain answers), provide the document processing agent 110 with training data, etc. As noted, communication between the station 152 and the admin interface 125 can be established based on any communication technology or protocol. To enhance security features, communications between the document processing agent 110 and the administrator station 152 may include authentication and/or encryption data (e.g., using symmetric or non-symmetric encryption keys provided to the document processing agent 110 and the administrator station 152). Via the communication link established between the administrator station 152 and the interfaces 120 and 125, the administrator provides information necessary for the document processing agent 110 to access the document library. For example, the administrator station can send a message providing the document processing agent 110 with a network address for the document library 160 (and/or identity of documents within that library that the agent 110 is to access and process). The administrator station can, in turn, receive an encryption key (e.g., a private symmetric key, or a public key corresponding to a private key used by the agent 110) that is to be used to encrypt content of documents that are to be transferred to the agent 110. The communication between the administrator station 152 and the admin interface 125 (or any of the other interfaces, such as interfaces 120 and 130, with which the administrator can communicate) can also be used to establish other configuration settings controlling the exchanges of data and information between the customer network 150a and the document processing agent 110, as will be described in greater detail below.
Once the AI agent has been provided with the location (e.g., represented as a network address) of the document library 160, and the communication features controlling the transmission of data between the customer network 150a and the agent 110, the agent 110 can begin receiving data transmissions of the documents from the repository (library) 160. The administrator station 152 can control the content sent, and perform some pre-transmission processing on the documents to be sent to the document processing agent 110, including removing sensitive content (e.g., private details), encrypting the content (e.g., using a public key corresponding to a private key at the document processing agent 110), authenticating the data to be transmitted, etc. The document processing agent 110 receives data transmitted from the customer network 150a via the server interface 120, and performs data pre-processing on the received data, including authentication and/or decryption of the data, format conversion (if needed), etc. The server interface 120 then passes the data corresponding to the documents sent from the document library 160 (subject to any pre-processing performed by the interface 120) to a document ingestion engine 126 that processes the received documents to transform (convert) them into a representation that allows the determination and generation of answers to queries provided by a user of the network 150a. Typically, prior to applying the transformation(s), the source document is segmented into portions (e.g., 200-word portions, or any other word-based segment), with the segmentation performed according to various rules for adjoining content from various parts of the documents into discrete segments. An example of a pre-processing (i.e., pre-transformation) rule is to construct segments using a sliding window of a fixed or variable length that combines one or more headings preceding the content captured by the sliding window, and thus creates a contextual association between one or more headings and the content captured by the window. Such a rule ensures that the transformation performed on a segment combines important contextual information with content located remotely (e.g., farther down the source document) from the segment being processed.
Having segmented the source document, and/or or performed other types of pre-processing (as will be described in greater detail below), the document ingestion engine 126 is configured to apply one or more types of transformations to the document segments to transform the segments into searchable segments (e.g., question-and-answer searchable segments). One type of transformation that can be applied to the segment is based on transforming the fixed-sized (or substantially fixed-sized) segments, typically comprising multiple words/tokens, into numerical vectors in order to implement a fast-search process. Such a search is typically a coarse search, in that it generally returns (in response to a query submitted by a user) a relatively high number of results (hits) because the search is based on matching vectors produced from input data comprising a relatively large number of words (tokens or features), and as a result the resolution achievable from such a transformation is lower than what can be achieved from transforming smaller segments. Thus, results based on coarse vector transformations might not provide as accurate representations of the textual meaning of the transformed content as other transformations applied on smaller segments. On the other hand, as the name suggests, the fast-search can be performed relatively quickly, and thus may be used to winnow the possible candidates of possible answers (to the submitted query) to a size or number that can then be more carefully searched (possibly through a search based on another type of transformation). Another transformation that may be applied by the ingestion engine is one for generating fine-detail vector transformations that are used to more narrowly pinpoint locations of answers with some text-segment (e.g., paragraphs) specific answer word sequences. Generally, document segments on which the fine-detail transformations are applied may be at a finer grain (resolution) than fast-search segments (which are generally of a fixed size, e.g., 200 words, and thus cannot typically pinpoint the exact location of an answer, if one exists, within the segment).
More specifically, a fast-search transformation (e.g., implemented through neural networks, filters, etc.) is applied to the segment to yield vectors with values that are based, and therefore are representative of, the content of the document segments. As will be discussed in greater detail below, several approaches may be applied by the document ingestion engine 126 to transform the data according to the fast-search transformation. In one example, the data representative of the content may be transformed into vector representations (e.g., fixed size vector, or variable size vectors). Thus, in such an example, the transform converts textual content into a vector of numerical values, which may or may not be associated with metadata (e.g., text-based metadata, providing additional information that can be used for further processing) or other contextual information. The resultant transformed vector can be representative of possible questions and answers that are associated with the input segment that was transformed. An example of a transformation that yields such vector-value representative of the content of the input (including contextual relationships) is the Bidirectional Encoder Representation from Transformers (BERT)
For the fine-detail transformation performed by the document ingestion engine 126, the source data (e.g., text-based portions segmented from a source document according to one or more rules or criteria, with the segmented portions typically being smaller in size than the source segments used for the fast-search transformation) is typically transformed into multiple vectorized (numerical) transformed content. The fine-detail transform may also be implemented according to BERT. The processing by the document ingestion engine 126 can include natural language pre-processing that determines at least some linguistically based information, such as detection and recording of locations of named entities (e.g., person and company names) in the document, expansion of structured data, such as tables, into searchable form of equivalent text, information conversion into knowledge representations (such as a predefined frame structure), etc. In some embodiments, the resultant fine-detail transformed data may be combined with the original content that is being transformed, along with derived or provided metadata (although such metadata is not critical, it can facilitate the performance of intelligent searching and question answering for a document). In some examples, the combination of the transformed content and the source segment can be further augmented with automatic questions that may be germane to the source segment, so that these generated questions are combined with the particular segment (or in a particular location in a full document that includes the entirety of the source content and the corresponding transformed content), or with a particular information field. When processing questions from a user, a similarity between the user's question and such automatically generated questions can be used to answer the user's question by returning the information (e.g., a pointer or actual user-understandable content).
With continued reference to
The DOM repository 140 is configured to (in conjunction with the document ingestion engine (also referred to as ingestion agent) 126 and/or the query processing module 136) store, manage, and search DOM records 142a-n. Content of a DOM record typically depends on the transformation performed by document ingestion engine 126. A DOM record can include data items associated with a particular source document or a source document portion. For example, one DOM record may be a collection of items that includes an original portion of a source document, metadata for that source document portion, contextual information associated with that source document portion, a corresponding coarse vector(s) resulting from a transformation applied to one or more fixed-sized (or substantially fixed-sized) segments of the original portion of the source document (to facilitate a fast-search process), a corresponding resultant fine-detail transformed content resulting from a fine-detail transformed (to facilitate a more accurate and refined textual search), etc. Thus, if the transformation resulted in a vector of values representative of the textual content of a segment, that vector is stored in the repository, possibly in association with metadata (added or embedded into the vector), and/or in association with the original content (in situations where the actual original text-content is preserved; in some embodiments, for security or privacy reasons, the source content may be discarded upon its ingestion, or may be available only at the customer's site). Metadata associated with the transformed content may include contextual information associated with the original source content, and location information that indicates the location or position of source content that resulted in the transformed content within the larger source document. Such location information can be provided in the form of pointer information pointing to a memory location (or memory offset location) for the source document stored in the customer network (i.e., so that when the pointer information is returned to a requesting user, it can be used to locate the memory location where the relevant content constituting an answer to the user's query can be found).
The transformed content (which may include several transformed content items, resulting from the various transformations applied to segmented content), metadata, and/or source content stored in the repository 140 together may define a unified record structure, in which each of the transformed content, metadata, and/or original source content is a field or a segment of the unified record structure. Individual records, when they correspond to discrete document segments of a larger source document, can be associated with each other (e.g., by arranging them sequentially or through logical or actual links/pointers) to define larger document portions (e.g., chapters for a particular document), or to define the entire original document that was segmented and ingested.
As further shown in
In embodiments in which the repository 140 includes multiple types of transformed source content, the search of the repository 140 may be implemented as a multi-pronged search. For example, because coarse numerical vector representation is generally more compact and easier to search (but may not as accurate as fine-detail transformed representations, whether achieved by a BERT-based transformation or some other transformation), a first prong of a search to determine an answer to a submitted query may be to convert the query data into coarse vector representation, and to use that first transformed query representation to search records in the repository 140 matching (e.g., according to some closeness criterion that may represent the distance, or difference, between the transformed vector query data and the transformed vector ingested content data) the coarse numerical-based transform of the query data. This type of initial searching may be referred to as fast-search. The results of the search may result in the identification of one or more answer candidates (e.g., identify 1000, or any other number, of possible segments that may contain an answer word sequence responsive to the query submitted by the user). The identified first batch of possible results can then be used to perform the second stage of the search by converting the query to a fine-detail transformed query and searching fine-detail transformed content associated with the search results identified in the first stage of the search process. This searching stage may be referred to as the detailed, or fine-grained, search. It is to be noted that, in some embodiments, the fast search may be used to identify the original portions of source content associated with the identified candidates, and those identified portions may then be transformed into fine-detail transform content. In such embodiments, the repository 140 does not need to maintain fine-detail transformed content, but rather the transformation of source content is done based on which portions have been identified by the fast-search as possibly containing an answer to the query. In alternative examples, searching for answer to a query may be performed directly on the entire fine-detail transformed content records without first identifying possible candidate portions of source content through a fast-search of fast-searched transformed content records.
Thus, in some embodiments, the query stack (e.g., the query processing module 136) is configured to transform the query data into transformed query data compatible with the transformed source content (e.g., compatible with one or more of the transformed content records in the DOM repository 140). For example, the fast-search-compatible transformation may be a coarse BERT-based transformation (e.g., using a learning engine implementing the same or similar trained learning model used to produce the searchable transformed content from the source data) that is applied to the entire query data (e.g., a natural language question) to produce a single vector result. The query processing module may, for example, launch a fast-search process in which it identifies one or more candidate portions in the transformed source content (with respective numerical vectors resulting from the coarse transformation) matching, according to a first criterion, the transformed query data. For example, the matching operation may be based on some closeness or similarity criterion corresponding to some computed distance metric between a computed vector transformed query data and various vector transformed content records in the repository 140. As described herein, in some embodiments, the transformed content may include vectors corresponding to possible questions that users may ask to which the source content provides a possible answer. The fast search may thus, in some embodiments, compare the transformed query result (generally a resultant vector record) to searchable vector records representative of possible questions that could be asked in relation to source content from which those searchable vectors were generated.
The query processing module 136 may be further configured to determine, from one or more fine-detail transformed content records corresponding to the one or more candidate portions identified based on their coarse transformed vectors, at least one fine-detail transformed content record matching, according to a second criterion (e.g., some other closeness or similarity metric), a fine-detail transformed data of the query data. Alternatively, in embodiments in which a fast-search is not performed, the query processing module 136 may be configured to identify one or more candidate portions in the transformed source content with respective fine-detail transformed content records matching, according to a second criterion, the transformed query data.
In some embodiments, the user query interface 130 and/or the query processing module may be coupled to a query cache 135 and a question generation unit (which may be part of the cache 135 or of the query processing module 136, or may be a separate unit). The query cache 135 stores, among other things, answers/contents corresponding to frequently asked questions. Such answers/contents may include content previously retrieved from the DOM documents (and/or from their corresponding raw source content) in response to previously submitted queries. Counters associated with such cached answers can track the frequency at which specific questions and answers have been submitted and/or retrieved. The cache 135 can also be configured to discard cached content that has not been requested within some reference (threshold) time interval. Content in the answer cache may also have been stored by the administrator (e.g., operating from a station, such as the station 152 via the admin interface 125) in anticipation of some likely questions that users of the customer system (network) 150a were expected to submit, or to override content that may have been retrieved from the DOM 140 (e.g., content that, based on subsequent feedback from users, was determined to be inaccurate or unresponsive to the query submitted). Thus, in some embodiments, the query stack is configured to determine whether received query data matches one of pre-determined questions (which may be stored in the answer cache), and to generate the output data based on one or more answer data records (possibly stored within the answer cache) in response to determining that the received query data matches one of the pre-determined questions. In some embodiments, the matching of query data to the past questions and associated answers stored in cache is performed by computing a score that is based on the combination of the questions and their answers, and ranking the computed scores to identify one or more likely matching candidates.
As noted, the query processing module may also include a question generation engine that can determine (e.g., based on a trained learning engine and/or using a repository of question data) follow-up or related questions to one or more questions submitted through the query data. In some embodiments, follow-up questions can be generated by paraphrasing the query submitted (e.g., transforming and/or normalizing the submitting query to modify the question submitted using, for example, a trained learning engine). In some embodiments, answer data determined for the submitted query (e.g., based on content retrieved from the DOM repository 140 via the query processing module 136) may be processed (by a separate module) to formulate further questions from the answer. Such derived questions can then be re-submitted to the query processing module to retrieve follow-up answers. This process can be iteratively repeated up to a pre-determined number of times. In some situations, the content stored in the DOM repository 140 may associate multiple questions (represented in whichever transformation format(s) that was applied during the document ingestion stage) with each processed segment of the source document (as noted, generation of transformed content may include, for each processed segment, data representative of questions associated with the processed segment, metadata, and content that may be provided in transformed format and/or the original source content). Thus, upon submission of a query (generally in transformed format computed, for example, according to a coarse-BERT or a fine-BERT type transformation), at least one DOM record/element will be identified. That search result may possibly be associated with multiple questions, including the question that may have resulted in a match between the identified resulted and the submitted query. One or more of the additional questions (i.e., other than the question that was matched to the query) may be used as a separate query to re-submit for searching to identify additional content that may be germane to the original query submitted by the user.
As further shown in
Generally, though, the query data is transformed (if it was not already transformed at the station 154a) by the query stack into transformed query data. The transformed data may provide the query in one or more transform formats that are compatible with the formatting of the transformed source content stored in the DOM repository 140. In some embodiments, the query data may also be used to generate one or more additional questions (e.g., follow-up questions, or questions related to the original query submitted by the user). In situations where an answer to the query is available from an answer cache, that answer itself may be used as a basis for generating further one or more questions that may be related to the cached answer(s). The query or the transformed query is used to search, via the query processing module 136, the DOM repository 140. As noted, the searching may be performed as a multi-pronged process according to multiple transformation formats used to store data in the DOM repository 140.
The output generated in response to a submitted query generally includes a pointer to the source content available at the customer network 150a. Because the data stored in the repository 140 is ingested based on source documents maintained at a document library available at the customer network, to which the user submitting the query has access, and because the source document might not have been stored in their original form at the document processing agent 110 (e.g., for security reasons, in order to protect sensitive data from being compromised), the output that is returned to the user does not require that actual answer data be sent back to the user. Instead, the pointer returned as the output of the query can identify the address or location of the answer with the appropriate document available to the user at the user's network 150. For example, in the illustrated example of
Additional details regarding the operation of the system 100 depicted in
As further shown in
The pointer to access the at least one portion of the source document may include one or more of, for example, document identifier identifying the source document in the repository of a plurality of source documents (i.e., in the document library 160), network address of the source document in the repository of a plurality of source documents, starting and end locations of the at least one portion of the source document within the source document, and/or a word span. In some examples, the generated output data may further include one or more of, for example, at least part of the source content corresponding to the at least one portion of the transformed content, and/or a summary of the source content corresponding to the at least one portion of the transformed content.
As noted, in some examples, the at least one portion of the transformed content may be stored in a repository of document object model (DOM) data items. A DOM data item corresponding to the at least one portion of the transformed content may include the pointer to access the at least one portion of the source document, and one or more of, for example, at least one transformed element representative of the at least one portion of the source content, the at least one portion of the source content, and/or contextual information associated with the at least one portion of the source content. The contextual information associated with the at least one portion of the source document may include feedback data from one or more users indicating responsiveness level of the output data to the question relating to the source content of the source document (e.g., user feedback of whether responses to a query include good or bad answers).
The transformed content in the repository 140 is generated according to ingestion processes that include, for example, segmenting source content, and then applying one or more different types of transformations on the segments (e.g., coarse/fixed vector transformation for fast search according to a BERT-based transformation, a fine-detail transformation according to a BERT-based transformation procedure, which may similar to or different than the transformation applied to the coarse transformation, and/or other types of transformations that are implemented through learning engines or through filtering operations). Thus, in some embodiments, the procedure 200 may further include receiving a copy of the source document at the local device, segmenting the copy of the source document into a plurality of document segments, and transforming each of the plurality of document segments into one or more respective transformed segments according to one or more transformations. An administrator may provide the document processing engine (e.g., via the admin interface 125 of
The one or more transformations may include one or more of, for example, a coarse linearization transform to generate numerical vectors representative of content of the plurality of document segments, and/or a fine-detail transformation to generate fine-detail transformed content records representative of the content of the plurality of document segments. Either of the coarse (also referred to as “fast-search”) or fine-detail transformation may be vectorized transformations (derived using, for example, neural networks), applied to one or more document segments, based on Bidirectional Encoder Representations from Transformers (BERT) processing. The coarse transformation is typically applied to fixed-sized (or substantially fixed sized) source data segments that are larger in size than the segments to which the fine-detail transformation is applied. In addition to the BERT-based transforms applied to source data and queries, additional types of transforms (whether utilized for fast-searching and/or detailed-searching) may be used (e.g., transformations to implement other NPL approaches such as Word2Vecot, GloVe, etc.)
In some embodiments, the procedure 200 may also include transforming the query data into transformed query data compatible with the transformed source content. In such embodiments, the procedure may additionally include implementing a fast-search by identifying one or more candidate portions in the transformed source content, produced through the coarse transformation process, matching, according to a first criterion, the transformed query data. Fine-grained searching may then be applied to the candidate portions identified by the fast-search. Thus, the procedure may further include determining, from one or more fine-detail transformed content records corresponding to the one or more candidate portions identified based on their respective coarse numerical vectors, at least one fine-detail transformed content record matching, according to a second criterion, a fine-detail transformed data of the query data.
In some examples, fine-grained searching is performed without first performing a fast-search. Thus, in such examples, the procedure may include identifying one or more candidate portions in the transformed source content with respective fine-detail transformed content records matching, according to a second criterion, the transformed query data.
In some examples, transforming the each of the plurality of the document segments may include transforming the each of the plurality of the document segments using a trained neural network implementing the one or more transformations. Segmenting the source content into the plurality of document segments may include segmenting the source content into the plurality of document segments according to one or more pre-processing rules. The segmenting can be performed so that context (e.g., heading information, font information, other metadata information) is added to the contents in the segments that are to be transformed. Thus, in such situations, segmenting the source content may include segmenting the source content into the plurality of document segments according to hierarchical rules semantically associating one portion of the source content with one or more other portions of the source content.
In some examples, the procedure may further include retrieving augmented data associated with the received source document, and segmenting a combination of the source document and the augmented data into the plurality of document segments. The data augmentation may be based on links found in the source document. Thus, retrieving the augmented data may include retrieving the augmented data based on links included in the source document.
The procedure 200 may further include transmitting the output data to the remote device. In some embodiments, additional questions may be generated. Thus, the procedure may also include identifying based on the received query data representative of the question one or more additional questions, from a repository of query data representative of a plurality of questions associated with source documents, to apply to the transformed content.
In some situations, answers to queries can be provided from a Q-A cache. Accordingly, the procedure 200 may include determining whether the received query data matches one of pre-determined questions, and, in response to determining that the received query data matches one of the pre-determined questions, generating the output data based on one or more answer data records associated with the matched one of the pre-determined questions.
As noted, document ingestion may be performed according to a fine-detail transformation (such as BERT, which in some embodiments may be a two-leg-BERT implementation as described in greater detail below). Thus, in some examples, the procedure 200 may further include partitioning the source content into one or more content portions, and transforming the one or more content portions according to a fine-detail transformation to generate fine-detail transformed content records representative of contents of the one or more content portions. The fine-detail transformation may include a transformation, applied to one or more document segments, based on Bidirectional Encoder Representations from Transformers (BERT) processing. Transforming the at least one of the one or more content portions may include transforming the at least one of the one or more content portions into transformed content based on one or more ground truth samples of question-and-answer pairs. In some embodiments, document summarization may be realized by picking up good sentences that are close to a summary sentence. Thus, in such embodiments, the procedure 200 may further include deriving a summary of the source document based on the transformed one or more content portions according to the fine-detail transformation. Deriving the summary may include identifying at least one of the transformed one or more content portions with a level of similarity to a pre-determined summary portion that exceeds a similarity threshold.
As noted, to facilitate Q-A content searching, source documents are ingested (e.g., via the document ingestion engine 126 of
In the example of
Ingestion of documents can be based on the specific source of data and/or on the desired or intended presentation of information (e.g., presentation of response data returned in reply to query data submitted by a user). For example, where the source of data (the content) is from some specialized application (Salesforce, Slack, etc.), the ingestion of the source content may be configured to perform specialized or dedicated pre-processing required for the specific source, e.g., convert chat data, or data arranged in specialized format records, such as records of Salesforce, into prose, or some other format more conducive to the transformations applied to segmented portions of the source content.
With respect to ingestion that is based on (or takes into account) the particular way the response data is to be presented, consider the following three examples of ways to achieve the data presentation. In a first example approach, data is presented according to an API-based methodology, where, for example, the answer/paragraph is included in addition to the location (such as page number or begin/end positions of the answer snippet) provided to a renderer of different format types (such as HTML, PDF, Word doc., etc.). The renderer can be implemented as a macro or plug-in/extension that allows for locating the answer snippet and paragraph in the document, and performing special processing of rendered segments e.g., by bolding or highlighting portions of the data, etc. of the segments. Another example approach for presenting response to data is to preserve, during the document processing phase (e.g., via the ingestion engines), screen shots of segments in the documents that are candidates for presentation (e.g., effectively, pre-rendering the output content). During a subsequent presentation of data identified as being responsive to a query, a client application can pick the most appropriate screenshot that holds the snippet/paragraph. In a third approach to present query results, after the appropriate segmentation for presentation is created, every segment of the processed documents, which may be available in different formats (e.g., as a Word doc, html, etc.), is converted to a PDF document format that includes the segment (with appropriate connections to the retrieval segments, where one-to-one mapping between segments are achieved and begin/end position of the answer snippet are passed through the API to a common PDF renderer) that allows for locating and highlighting the answer snippet.
One example of a pre-processing procedure is the segmentation of source content for a source document into multiple document segments. Such segmentation can be performed according to hierarchical rules semantically associating one portion of the source document with one or more other portions of the source content. For example, a sliding window of a fixed or variable size (e.g., 200 words) can be applied to the source content to generate manageable-sized segments on which to apply content transforms. However, when segmented into small chunks, the content segments may lose important contextual information that otherwise would have been available for a larger size segment. For example, a passage in the middle of a section of a document may, in isolation, not include important contextual information such as the section heading, location of the passage relative to earlier passages in the section, font sizes associated with other passages not captured by a particular segment (e.g., whether the present passage is a footnote), etc. Therefore, in some embodiments, contextual information (e.g., section heading, chapter heading, document title, location, font type and size, etc.) may be combined with one or more of the document segments. This pre-processing procedure is illustrated in
In some examples, to simplify the segmentation process (so as to facilitate more efficient searching and retrieval), the source documents may be segmented to create overlap between the sequential document segment (not including the contextual information that is separately added to each segment). Thus, for example, in situations where a segment is created by a window of some particular size (constant or variable), the window may be shifted from one position to the following position buy some pre-determined fraction of the window size (e.g., ¾, which for a 200-word window would be 150 words). As a result of the fractional shifting, transformations (e.g., linear vectorization of BERT-based transformations) applied to overlapped segments results in some correlation between the segments, which can preserve relevancy between consecutive segments for subsequent Q-A searching. In some embodiments, heading information (and other contextual information) may be added directly to partitioned segments. Alternatively, heading and contextual information may either be transformed into vectors that are then added to the vectors resulting from transformation operations applied to the content extracted by the sliding window, or may be combined with the content extracted by the window before the transformation is applied to the resultant combined data. By associating neighboring segments with each other (e.g., through fractional shifting of the window over a document to form the segments), identification of relevant paragraphs (responsive to submitted queries), for the retrieval and presentation processing for top paragraphs and associated answer snippets, is improved.
Another pre-process that can be applied during segmentation of the source document relates to the handling of table information (i.e., when the original content is arranged in a table or grid). This pre-processing is used expand structured data arranged in tables (or other types of data structures) into searchable form such as equivalent text. For example, upon identifying a portion of the source document as being a multi-cell table, substitute portions are generated to replace the multi-cell table, with each of the multiple substitute portions including a respective sub-portion content data and contextual information associated with the multi-cell table. Additional examples of pre-processes include a procedure for associating contextual information with one or more portions of the source document based on, for example, a) information provided by a user in response to one or more questions relating to the source document that are presented to the user, and/or b) based on one or more ground truth samples of question-and-answer pairs.
In some examples, contextual information might not be explicitly included with a segment, but instead may need to be discovered, and included with document segments as augmented information (in this case, augmented contextual information). For example, entity discovery (determining identity of relevant entities referenced in the document) can be used to help speed up the search (fast-match (FM) searching, or detailed match (DM) searching) during inferencing, and to improve searching accuracy and generate an improved schema.
Consider the following example implementations:
Information about a specific entity (or entities) relevant to a user's search can also be used to generate more accurate additional questions (e.g., to determine different ways to paraphrase the input query so that additional possible question-answer pairs can be generated), and also to provide additional context that can be used to search the repository of data (be it DOM objects in transformed form, or user-readable data formatting).
Entity contextual information may also include information about specific items (e.g., different product brands) and elements that provide additional contextual information to refine the search and improve output results. For example, consider a situation in which a user is interacting with an augmented reality system equipped with cameras (and/or other types of sensors). In such a system, one or more of the cameras will be pointing at location that the user is looking at. Information in the scene captured by the sensor device (e.g., image data, which can be processed by, for example, a learning machine to identify objects and items appearing in the scene) can be used to provide contextual information to a query concomitantly initiated by the user. For instance, if the user looks down (and a camera of the augmented reality system similarly follows the direction and orientation of the user's head to point at the scene being viewed by the user), sees a MagSafe charger (for wireless charging) for his/her phone, and asks “how do I charge my phone?,” a Q-A system (e.g., based on the implementations described herein) will identify different answers for this questions (resulting from a search of the DOM repository) than would be identified if the user were looking down and seeing a car. That is, the sensor of the augmented reality system is used to determine (or discover) contextual information (e.g., proximity of the user to a MagSafe charger vs. proximity to a car), and perform a search that relies, in part, on that determined contextual information.
In some embodiments, the orientation, positioning, and or location (as may be determined based on positioning techniques using satellite or ground-based signal analysis) of the sensor device (the camera, in this case) can itself provide important contextual information that is germane to a search of the repository data. For Example, pointing the camera in a downward direction can imply that the information being sought via a query relates to objects that are situated close to the ground. In another example, location of the sensor device can be used to limit the search to answers that have relevance to the particular geographic location of the sensor device (e.g., to determine details related to a specific conference room where the user is located). Thus, a query such as “how do I turn on the video conference camera?” can be modified (or be restricted) to search answers (e.g., from relevant manuals or other source documents stored by the company) for the video camera(s) located within the particular conference room where the user posing the query is located.
Another example where augmented reality systems (or other types of systems equipped with sensors) can be used in conjunction with the document processing (e.g., Q-A type processing) implementations described herein involves situations where factory workers, who may be fitted with streaming bodycams (or hard-hat-cams), can pose queries questions that are modified by contextual information extracted from the captured video stream. A user may, in one situation, ask information about functionality or operation of a “machine,” or information about a certain “product.” The captured image or video by the device carried by the user can identify the particular brand or model of the machine, and, when the user is asking for some specific information about the operation of the machine, the specific model identified through the augmented reality sensor can be used to restrict the search to documents (e.g., user manuals) relevant to the specific machine model identified. Thus, streaming by a camera when used in an augmented reality system adapted to assist factory workers can be used to modify queries (e.g., seeking information about “a machine”) to account for the specific machinery identified according to the video streams generated by a camera (used in conjunction with a learning machine to identify objects and items in the scenery). In another example related to the factory-worker (or technician) scenario, a user may pose a query (e.g., through a voice-based interface, such as an AI assistant app operating on a mobile device carried by the user) asking about the connectivity of a wiring harness. The query may be modified (or restricted) to search answers that may be specific to a wiring harness appearing in a captured image (or video) of the scene, from which the specific model or type of the harness can be identified.
Another example where captured image-based data can be used in the course of contextual discovery is when the scenery includes recognizable codes (such as QR codes, barcodes, etc.) that can be decoded to extract meaningful contextual information therefrom. For example, in the above wiring harness example, the wiring harness may include a label with a QR code or a barcode that can be automatically decoded upon being captured by the image-capture device carried by the user. Queries then posed by the user in relation to the wiring harness will be modified (or restricted in some way) so that the answer(s) obtained are relevant to the QR or barcode identified during the context discovery.
It is to be noted that some of the example systems (e.g., augmented reality systems) described herein can be implemented using augment reality goggles (glasses), while other systems can be implemented using cameras installed on smartphone that the user moves to point the camera in the direction of the relevant scenery. Some embodiments of such phone-based augmented reality system may also include an Artificial Intelligence (AI) assistant app (e.g., Ski, Alexa, Cortana, etc.) through which the user may provide his/her queries that are modified based on contextual information determined from the augmented reality system. It is also to be noted that other types of mix-mode input sources to formulate queries (in the course of searching Q-A data repository) can be used, that may combine inputs from one or more of text-entry sources, voice-capturing sources, image-capturing sources, etc.
In some embodiments, document processing (e.g., segmentation) can be performed as two separate tasks. In one processing task, the source document is properly segmented and organized into small chunks, e.g., paragraphs, with additional augmentations (e.g., the vector sequence that represents the heading of a section can be appended to the vectors of every paragraph in that section). These augmentations are used to improve the retrieval accuracy. In a parallel task, a document is segmented in the most appropriate way for presentation purposes. The two different resultant segmentation outputs need to be associated with each other such that when, during retrieval processing, the top paragraphs and associated answer snippets are identified, but what is presented to the user are the presentation contents (rather than the identified answer snippets) associated with the identified answer snippets. In other words, the system can ingest a particular passage to facilitate searching operations, and separately ingest that particular passage to facilitate presentation operations. In this example, upon identifying the passage as a result of matching a query to the searchable ingested content, the presentation content associated with the identified passage is outputted.
Having segmented a source document into multiple segments, each segment may be provided to one or more content transforms (or transformers) 330a-m that transform the segment (content, and optionally the contextual information, although in some embodiments the contextual information may be preserved without transforming it) into a resultant transformed content that is associated with question(s) and answer(s) related to the original content of the respective segments. In the example of
As noted above, an example of transforms that may be applied is the fast search (also referred to as a fast-match, or a coarse search) transform that is based on transforming fixed-sized (and typically large) segments of input data into vectors (the vectors too may be, but do not necessarily have to be, of uniform dimensions). The resultant transformed vectors can be representative of possible questions and answers that are associated with the input segment that was transformed. The resultant vectors generally provide a starting point to narrow the number of possible document objects that need to be searched more thoroughly (e.g., using content transformed according to another, more fine-grained, transforms). For example, upon searching the transformed content repository (e.g., the DOM repository 140) based on a match between the fast-search transform results and query data converted into a representation compatible with the fast-search transformed content, the resultant search can result in, for example, 1000 potential candidates (or any other number of candidates). More refined content matching can then be performed on transformed content objects that correspond to the candidates identified by searching the fast-search transform content. The fast-search (coarse) transformation may be implemented according to the BERT-approach. Another transform, illustrated as being performed by transform unit/module 330b in
Under the BERT approach, when a query is received, the relevant sequences in the documents can be identified quickly (possibly from a set of objects that may have been earlier identified using, for example fast-search processing) by identifying a part of a document (e.g., a paragraph) that may contain the answer, and identifying the span of words in that part of the document that contains the specific answer. In some examples, under the BERT approach the question and the answer are concatenated (tokenized for example using WordPiece embeddings, with suitable markers separating the question and the answer) and processed together in a self-attention-based network. The output of the network indicates a score for each possible starting position for the answer and a score for each possible ending position for the answer, with the overall score for a span of the answer being the sum of the corresponding start and end positions of the answer. That is, a self-attention method is used where embedded vectors of a paragraph and a query are mixed together through many layers followed by a decision-maker layer and segmenter logic to provide an efficient method to determine if a question is answerable by a paragraph, and if so, determine where exactly the span of the answer lies in the paragraph.
In the BERT-based approach, a network may first be trained on a masked language model task in which a word is omitted from the input, and predicted by the network by an output layer that provides a probability distribution over words of the vocabulary. Having trained the network on the masked language model task, the output layer is removed, and in the case of the question answering task, a layer is added to yield the start, end, and confidence outputs, and the network is further trained (e.g., fine-tuned, transfer learning) on supervised training data for the target domain (e.g., using Stanford Question Answering Dataset, or SQuAD). Having trained the network for question answering for the target domain, further training may be used to adapt the network to a new domain. Another training strategy used for BERT is the next-sentence prediction, in which the learning engine is trained to determine which of two input segments (e.g., such segments may be neighboring sentences of a text-source) is the first of the two segments. When training the model, both the masked-language and next-sentence training procedures may be combined by using an optimization procedure that seeks to minimize a combined loss function. Alternatively, or additionally, other training strategies (to achieve context recognition/understanding) may be used separately, or in conjunction with, one of the aforementioned training strategies for BERT.
In example embodiments based on the BERT approach, an implementation, referred to as a Two-Leg BERT approach, may be used in which much of the processing of a query is separated from the processing of parts of a document (e.g., paragraphs) in which answers to the query may be found. Generally, in the two-leg-BERT approach, the neural network architecture has two “legs”, with one leg for processing the query, and one for processing the paragraph, and the outputs of the two legs are sequences of embeddings/encodings of the words of the query and the words of the paragraph. These sequences are passed to a question-answering network. A particular way this approach is used is to precompute the BERT embedding sequences for paragraphs, and complete the question-answering computation when the query is available. Advantageously, because much of the processing of the paragraphs is performed before a query is received, a response to a query may be computed with less delay as compared to using a network in which the query and each paragraph are concatenated in turn and processed together. The paragraphs are generally much longer than the queries (e.g., 200-300 words versus 6-10 words) and therefore the pre-processing is particularly effective. When successive queries are applied against the same paragraph, the overall amount of computation may be reduced because the output of the paragraph leg may be reused for each query. The low latency and reduced total computation can also be advantageous in a server-based solution. As noted, in the implementations described herein, the BERT-based processing of the source documents produces transformed content that is typically stored in a repository (such as the DOM repository 140 of
In some embodiments, the BERT-based transformers (e.g., used for the fast, coarse, transformation, and/or for the fine-detail transformation) may be implemented according to encoder-based configuration. For example, a BERT-based transformer structure may include multiple stacked encoder cells, with the input encoder cell receiving and processing the entirety of an input sequence (e.g., a sentence). By processing the entirety of an input sentence, a BERT-based implementation can process and learn contextual relations between individual portions (e.g., words in the input sequence). An encoder layer may be realized with a one or more self-attention heads (e.g., configured to determine relationships between different portions, e.g., words in a sentence, of the input data), followed by a feedforward network. The outputs of different layers in an encoder implementation may be directed to normalization layers to properly configured resultant output for further processing by subsequent layers.
It is to be noted that, in some embodiments, the fast-search vector transformation (transforming a segment into a compact-sized numerical vector) may be applied to a tokenized version of the text (e.g., some transformation, such as transformations achieved through the BERT process, may have already been performed to produce an intermediary (e.g., tokenized) content, to which the fast-search transform is then applied).
The transform modules (fast-search, BERT-based, or any other type of transform) may be implemented through neural networks that have been pre-trained to produce transformed content associated with question-answer pairs. Other transform implementations may be realized using filters and algorithmic transforms. Training of neural network implementations may be achieved with a large training samples of question-answer ground truths that may be publicly available, or may have been internally/privately developed by the customer using the system 100 to manage its document library.
In implementations based on neural networks, different types of neural network architectures, configuration, and/or implementation approaches may be used. Examples neural networks that may be used include convolutional neural network (CNN), feed-forward neural networks, recurrent neural networks (RNN), etc. Feed-forward networks include one or more layers of nodes (“neurons” or “learning elements”) with connections to one or more portions of the input data. In a feedforward network, the connectivity of the inputs and layers of nodes is such that input data and intermediate data propagate in a forward direction towards the network's output. There are typically no feedback loops or cycles in the configuration/structure of the feed-forward network. Convolutional layers allow a network to efficiently learn features by applying the same learned transformation(s) to subsections of the data.
The neural networks (and other network configurations and implementations for realizing the various procedures and operations described herein) can be implemented on any computing platform, including computing platforms that include one or more microprocessors, microcontrollers, and/or digital signal processors that provide processing functionality, as well as other computation and control functionality. The computing platform can include one or more CPU's, one or more graphics processing units (GPU's, such as NVIDIA GPU's, which can be programmed according to, for example, a CUDA C platform), and may also include special purpose logic circuitry, e.g., an FPGA (field programmable gate array), an ASIC (application-specific integrated circuit), a DSP processor, an accelerated processing unit (APU), an application processor, customized dedicated circuitry, etc., to implement, at least in part, the processes and functionality for the neural network, processes, and methods described herein. The computing platforms used to implement the neural networks typically also include memory for storing data and software instructions for executing programmed functionality within the device. Generally speaking, a computer accessible storage medium may include any non-transitory storage media accessible by a computer during use to provide instructions and/or data to the computer. For example, a computer accessible storage medium may include storage media such as magnetic or optical disks and semiconductor (solid-state) memories, DRAM, SRAM, etc.
The various learning processes implemented through use of the neural networks described herein may be configured or programmed using TensorFlow (an open-source software library used for machine learning applications such as neural networks). Other programming platforms that can be employed include keras (an open-source neural network library) building blocks, NumPy (an open-source programming library useful for realizing modules to process arrays) building blocks, etc.
With reference next to
In some embodiments, applying the one or more pre-processes may include segmenting the source document into a plurality of document segments. Segmenting the source document may include segmenting the source document into the plurality of document segments according to hierarchical rules semantically associating one portion of the source document with one or more other portions of the source content. In such embodiments, segmenting the source document according to hierarchical rules may include combining (including), in a particular document segment, content of a particular document portion and section heading content located in the source document ahead of a location of the particular document portion, with the section heading content being determined to be associated with the content of the particular document portion.
In some examples, applying the one or more pre-processes to the source document may include identifying a portion of the source document comprising multiple sub-portions arranged in a multi-cell table, and generating multiple substitute portions to replace the multi-cell table, with each of the multiple substitute portions comprising a respective sub-portion content data and contextual information associated with the multi-cell table. In some other examples, applying the one or more pre-processes to the source document may include associating contextual information with one or more portions of the source document based on information provided by a user in response to one or more questions relating to the source document that are presented to the user. Additionally or alternatively, applying the one or more pre-processes to the source document may include associating question-and-answer contextual information relating to a particular portion of the source document based on one or more ground truth samples of question-and-answer pairs.
Applying the one or more pre-processes to the source document may also include determining relative importance value for a particular portion of the source document based on one or more of, for example, location of the particular portion relative to locations of one or more other portions of the source document, relative font size of the particular portion, structure and organization of the source document, document type (PDF vs. Word documents) of the source document, etc.
As noted, in some variations, pre-processing can include the determination of important contextual information (e.g., entity, task, subject matter) that can be used to tag segments, and, subsequently, upon submission of a query, to eliminate from searching those searchable document segments whose tags are inconsistent or contradictory to the tags/descriptors associated with the query data. That is, query data is processed to derive contextual data in the form of tags or descriptors (e.g., using rules, heuristics, classification engines (that may be based on learning engine implementations), and those tags/descriptors may be matched up to document segments (e.g., in transformed form) that have matching/consistent tags/descriptors. Thus, in such variations, applying the one or more pre-processes may include segmenting the source document into a plurality of document segments. In these examples, the procedure 400 may further include for at least one segment of the plurality of document segments, determining at least one segment descriptor comprising one or more of, for example, at least one entity associated with the at least one segment, at least one task associated with at least one segment, and/or subject matter descriptor associated with at least one segment, and tagging the at least one segment with the at least one descriptor. The procedure may also include receiving, at the local device, query data representative of a question relating to source content of the source document, determining at least one query descriptor associated with the query data, the at least one query descriptor comprising one or more of, for example, at least one entity associated with the at query data, at least one task associated with the query data, and/or subject matter descriptor associated with the query, and searching a response to the query data from one or more of the plurality of document segments with segment descriptors matching the at least one query descriptor. As noted, searchable segments with tags that are inconsistent with the query tags/descriptors may be excluded/eliminated from the searching process. In another example, the method may further include receiving, at the local device, query data representative of a question from a user relating to the content of the source document, determining contextual information based on sensor data obtained by a sensor device associated with the user, and searching a response to the query data from one or more of the plurality of document segments with segment descriptors matching the determined contextual information. In such embodiments, determining the contextual information may include determining an item or location identifiable (e.g., using a learning machine configured to detect and identify objects appearing in an image) from image data captured by a camera device used by the user, and searching the response to the query data may include searching the response to the query data from one or more of the plurality of document segments based, at least in part, on the determined item or location identifiable from the image data. It will be appreciated that contextual information can also be obtained from other sensor devices associated with the user, including audio sensors, wireless communication sensors (e.g., wireless transceivers, satellite positioning receivers, etc.) that are included with a mobile device used by the user submitting the query. As noted, in some examples, the sensors devices may be part of an augmented reality system that includes a streaming image-capture sensor. It is to be noted that the determination of contextual information to facilitate the searching executed in response to a received query can be performed not only with respect to the procedure 400, but also with respect to any of the other procedures and methods of the present disclosure.
With continued reference to
As described herein, to reduce search time, improve accuracy, and be able to override bad answers that may have been obtained through searching of the transformed content on the DOM repository, an input cache, such as the cache 135 schematically depicted in
As noted, in some embodiments, a customer network (such as the customer network 150a of
To facilitate access to knowledge specific to respective applications operated by the customer, widgets (e.g., implemented based, for example, on Java programming language, to cause the invocation of a query process or window) may be implemented to provide query interfaces through which users can enter questions (e.g., “how to” questions) pertaining to that specific application (this is referred to as “in-application searches”). For example, consider
To further improve on the efficiency and accuracy of determining responses to queries, each customer network application/platform that constitutes a separate a source of data to be ingested may be associated with a separate, dedicated, cache structure to store data that originated from that application (for each such cache, the criteria for cache storage may be similar to those discussed with respect to the cache 135, namely, based on frequency at which questions and retrieved responses are sought, based on whether particular responses have been flagged as satisfactory, based on whether expert users have compiled responses to specific questions, etc.) Although different applications may be associated with dedicated caches, it will be appreciated that the cache can be consolidated at a physical location(s) that is logically partitioned to serve specific applications, or that a single cache (whether at a single physical location or at multiple distributed locations) can be associated (e.g., using an identifier, flag, or record field) with one or more of the multiple applications used by the customer network. Thus, if the customer network 150a uses the Slack application (or any other chat, information sharing, or collaborative application), data generated via that application may be processed and preserved at the DOM repository 140, with at least some of the information stored at a dedicated cache for that application, or at a cache section (of a global cache, such as the cache 135 of
A user may be authorized to access and use one or more of the various applications available to users of the customer network 750 (some of the applications may be running on local servers, while one or more of the applications may be cloud-based applications). Some or all of the data generated, or otherwise provided to the applications a-m may be forwarded (e.g., as may be facilitated by network administrators) to a document processing agent 710, which may be similar in configuration and/or implementation to the agent 110 of
With reference next to
The search of the question-answer cache (also referred to as a “question store) may be based on meaning representation and using question similarity (paraphrase) processing. Thus, in some examples, determining whether the one or more pre-determined questions stored in a question-answer cache maintained at the local device matches the query data according to one or more matching criteria may include determining linguistic meaning of the query data, determining a level of meaning similarity between the query data and respective meaning of at least some of the pre-determined questions in the question-answer cache, and determining that at least one of the one or more pre-determined questions is a match to the query data when the level of meaning similarity between the query data and the respective meaning for the at least one of the one or more pre-determined questions exceeds a pre-determined similarity threshold. In some embodiments, matching the query data to cache data could also be implemented through exact OR operation based on a process similar the BERT-based model. In such embodiments, a vector of the query is matched to a combination of vectors of the frequently asked question and the vectors of the answers that have previously been found. Thus, in such embodiments, a new query may be matched to existing old questions in the cache (which as noted, may be a global cache to store Q-A pairs from all content sources of an entity, or dedicated caches associated with individual content sources generated using different applications) and/or to existing answers. The matching process includes determining the closeness or similarity of the current query data to questions, to answers, and/or to question-answer pairs. Matching of a new question (corresponding to the current query) to existing questions can be dealt with as a paraphrasing problem, whereas matching of the new question to an old answer can be thought of as reading-comprehension matching problem. A combination of the two matching steps (the matching of the new question to existing questions, and of the new question to existing answers) can thus lead to improved performance.
With continued reference to
User feedback to the query results delivered to a user (responsive to a query submitted by that user) may be used to cache a Q-A pair that the user has indicated to be a good answer to the query he/she has submitted. Thus, the procedure 500 may further include storing in the question-answer cache a data item representative of a pair of a particular question and a corresponding answer in response to receiving an indication, from a user, that a portion of the data item representative of the corresponding answer represents a satisfactory response to the particular question.
In some examples, when a determination is made that an answer delivered in response to a query (e.g., whether the answer was obtained as a result of searching the cache 135 or the DOM repository 140) is incorrect or unsatisfactory (i.e., it is a “bad” answer), a correct answer can be provided (e.g., by a user of the customer network) and stored in the cache to override the incorrect answer. Thus, the procedure 500 may further include obtaining, at the local device, replacement answer data to replace an initial answer data determined at the local device in response to a particular question, with the initial answer determined by a user to be an unsatisfactory response to the particular question, and storing in the question-answer cache a data item representative of a pairing of the particular question and the replacement answer.
In some embodiments, the main document storage (e.g., the DOM repository 140 of
Implementations described herein are also configured to generate follow-up questions recommendations (e.g., by the query processing module 136 or the user query interface 130 illustrated in
In some embodiments, the systems described herein may generate additional recommended question as part of a question-augmentation procedure. In such embodiments, for a QA system (such as the system 100 of
Thus, in embodiments relating to generation of augmented questions, the procedure 500 may further include generating one or more augmented questions based on the question corresponding to the query data. Generating the one or more augmented questions may include revising the question in the query data based on one or more of, for example, a) changing order of wording in the question corresponding to the query data, and/or b) substituting one or more words in the question with substitute words identified from a synonym data record (e.g., synonym file) based on a determination that the substitute words are synonyms or equivalent words to the one or more words in the question. The procedure may further include searching at the local device a documents storage, comprising a plurality of document objects of transformed content corresponding to one or more source documents, to identify one or more of the plurality of document objects according to the one or more augmented questions.
Implementations described herein are additionally configured to determine if a received query is a command (rather than a question). Thus, in such implementations, the procedure 500 may further include determining that the query data matches one of pre-determined commands associated with respective processes performed, at the local device, with respect to the source document, and executing the pre-determined process associated with the one of the pre-determined commands determined to match the query data. In some examples, determining that the query data matches one of pre-determined commands may include determining that the query data matches a summary command configured to cause generation of a summary of at least a portion of the source document.
The implementations described herein may also be used to curate data. Particularly, a question-and-answer (Q-A) data item (e.g., a response to a query submitted by a user) is selected (e.g., based on metadata) and is sent to one of several experts for data curation (refinement) of the Q-A data item. The expert can be selected based on information about the available experts by, for example, matching subject matter of the Q-A data item to the right expert, and/or based on other criteria. The expert edits the item, adding some additional data (to capture institutionalized/tribal knowledge), and sends the refined answer back to archiving (e.g., in a knowledge cache, which may be part of the cache 135 of
The metadata used to determine whether to send a Q-A item to an expert for curation may include history information representative of historical information about formation and subsequent access of the Q-A item. Examples of such history information include frequency at which Q-A item has been accessed by users, feedback from users indicating whether an answer is good or bad, confidence level of responsiveness of an answer to a question, authorship of the source document from which the Q-A data item has been mined, etc. The confidence level of responsiveness (one of the types of history information based on which the determination is made of whether to send a data item to an expert for refinement) can be based on a ranking of search results performed on ingested documents, e.g., confidence level will depend on how many “hits” were generated, and how close is a question associated with an identified document portion to an original query submitted by a user. The historical information for a Q-A data item can be compared to respective threshold values to decide whether or not to send the Q-A data item to a selected expert. If, for example, the confidence level associated with the highest ranked answer is below some pre-determined threshold, a rule-based mechanism implemented by the query module nay determine that the query (and, optionally, one or more of the responses identified by running the query on the DOM repository) needs to be directed for further review by an expert (e.g., an experienced user). A different example of a rule or process to handle search results may include considering the confidence level associated with the top three (3), or any other number of top-ranked answers, and determining that a query needs to be sent to an expert when all those top-ranked answers are below some pre-determined threshold (which may be the same, or different, than the threshold used to evaluate just the top-ranked answer). Yet another example of a rule to handle search results may be to use a combination of a confidence level and feedback by a user (e.g., the confidence level of response(s) provided to a user is/are below a third pre-determined threshold, and the user indicates the responses provided are unsatisfactory). Conversely, in this example, if a user indicated the response(s) to be acceptable, no further curation would be required even if the confidence level for the response is below the third pre-determined threshold. Other rules or processes for handling the search results (e.g., based on the historical information) may be used. The rules and processes may be performed according to some order or rule hierarchy implemented to handle responses (e.g., multiple rules may be applied in a certain order to a given set of query responses, until conditions imposed by one or more of the rules or processes are satisfied).
When an expert returns a curated (refined) Q-A data item, the curated Q-A data item is cached (e.g., stored in the cache 135). The curated data item is communicated to the agent 110 via the user query interface 130, or via a dedicated expert interface 132, which may be implemented similarly to any of the other interfaces of the system 100, but may only be accessed by the selected expert (e.g., an expert at the station 154b of
If on a subsequent query the query data matches the question portion of a curated Q-A item stored in the cache, the answer portion of the curated Q-A item is used to generate output data deliverable to the user that submitted the query. If, however, on the subsequent query the query data does not match any of the question portions of the Q-A data items in cache (whether those items were curated or stored as a result of some of the other processes described herein), the query data is used to perform a regular search of the DOM repository to obtain a possible response to the query. As noted, in some embodiments, the matching of query data to cache items may be implemented based on a matching procedure that matches the query to the question in the cache, and to the answer items in the cache (e.g., based on a BERT-based process).
Thus, with reference to
The question-and-answer history information (which may be part of the metadata associated with the Q-A data item) may include one or more of, for example, frequency at which the associated question-and-answer data item has been accessed, level of satisfaction determined based on feedback by one or more users regarding the question-and-answer data item, a confidence level of responsiveness of an answer portion to a question portion of the question-and-answer data item, and/or authorship information associated with the question-and-answer data item.
In some examples, the procedure may additionally include deriving the confidence level of the responsiveness of the answer portion to the question portion according to search results obtained from searching a plurality of document objects of transformed content corresponding to one or more source documents, to identify one or more of the plurality of document objects according to question data representative of the question portion of the question-and-answer data item. In such examples, the confidence level is based on what was found using a search of the DOM repository. Thus, deriving the confidence level may include deriving the confidence level of the responsiveness of the answer portion to the question portion based on a number of the identified one or more plurality of document objects, and on closeness level of the question data to respective question portions associated with the identified one or more of the plurality of document objects. In some embodiments, determining based on the question-and-answer history information whether to perform refinement operations on the question-and-answer data item may include determining whether to refine the question-and-answer data item based on one or more of, for example, i) the frequency at which the associated question-and-answer data item has been accessed exceeding a frequency threshold, ii) the level of satisfaction determined based on the feedback by the one or more users regarding the question-and-answer data item being below a satisfaction threshold, iii) the confidence level of the responsiveness of the answer portion to the question portion exceeding a responsiveness threshold, and/or iv) the authorship information matching one or more pre-identified names.
With continued reference to
The Question-and-answer data item may include one or more of, for example, a pointer to access from a remote device at least a portion of the source document corresponding to an answer to the question-and-answer pair, at least part of the source content, and/or a summary of the at least the portion of the source content. In some embodiments, the procedure may include storing a refined (curated) question-and-answer pair in a question-answer cache at the data processing agent. Thus, the procedure 600 may further include receiving, at the local device, a refined question-and-answer data item derived from the question-and-answer data item communicated to the destination device, and storing the refined question-and-answer data item at a question-answer cache maintained at the local device.
On subsequent queries that match a question element for a curated question-and-answer pair stored in the cache, the curated answer element stored in cache for the curated question-and-answer pair is sent back to the querying user. Thus, the procedure may further include receiving, at the local device, from a remote device (e.g., the station 154a where a regular user is submitting a query) query data representative of a subsequent question relating to the source content of the source document, determining whether the query data matches a question portion of the refined question-and-answer data item, stored in the question-answer cache, according to one or more matching criteria. In response to a determination that the question portion of the refined question-and-answer data item matches the query data received from the remote device, obtaining from the question-answer cache at least one answer portion of the refined question-and-answer data item, and transmitting the at least one answer portion to the remote device.
On the other hand, if the query data submitted does not match any question element in the cache, the query is submitted as a search request to the document repository (e.g., the DOM repository 140 of the system 100 in
Additional embodiments of the implementations described herein may include some of the following features.
The document processing agent 110 may be configured to implement document library maintenance, cleanup, and administration features. For example, the agent 110 may be configured to detect inconsistent and duplicate answers/documents. During use (e.g., with admin reviewing suspect answers from logs), the system may be configured to allow either the admin or the user to remove the error documents. In some examples, the agent 110 may implement a detection feature that can issue warnings admin if a pair of paragraphs stored in the document library in different documents are too close or identical (e.g., in terms of their content), etc.
Thus, in some embodiments, a method is provided that includes applying question-and-answer processing on multiple source documents from one or more document libraries to generate respective resultant document records, determining level of similarity between a first one and a second one of the resultant document records, and managing the one or more document libraries based on the determined level of similarity between the first one and the second one of the resultant document records.
Example embodiments include one or more of the following features. Managing the one or more document libraries may include determining, based on the determined level of similarity, that the first one and the second of the resultant document records are inconsistent with each other, or are substantial duplicates of each other. The managing may further include removing from the one or more document libraries one of a first source document corresponding to the first one of the resultant document records, or a second source document corresponding to the second one of the resultant document records, in response to a determination that the first one and the second one of the resultant document records are either inconsistent or substantial duplicates of each other.
Determining the level of similarity between the first one and the second one of the resultant document records may include determining the level of similarity based on user input responsive to output data, generated from the question-and-answer processing, provided to the user. Determining the level of similarity between the first one and the second one of the resultant document records may include determining the level of similarity, during document ingestion to create questions-and-answer objects, based on one or more of, for example, one or more hypothetical library questions applied to the multiple source documents, or determining a question from the first one of the resultant document records and using the determined question to evaluate content of the second one of the resultant document records. The method may further include providing output indicating similarity between a first source document, corresponding to the first one of the resultant document records, and a second source document, corresponding to the second one of the resultant document records, in response to a determination that the level of similarity between the first one and the second one of the resultant document records exceeds a similarity threshold.
Another set of features that may be included with the document processing implementations described herein relate to conversational reading. For this set of features, a document (e.g., article) may be downloaded with pointers to related content. The related content can be ingested to augment it for question answering (e.g., references on a Wiki page). The library of documents can be enhanced to include the related content for question answering dynamic extension document content (also known as conversational reading).
In some implementations, related complementary content (to expand on the original) may be obtained by: (a) using URL's in the original document to retrieve the content of those URL's into the knowledgebase, and/or (b) perform an entity and co-reference (CoRef) resolution (a natural language processing procedure for identifying expressions and terms that may refer to a particular name or term in the original content source) performed on the source document to find a set of important entities, and to run search tasks to find documents in large repositories (the entire Internet, private networks containing data repositories, etc.) that are relevant to the original document. Documents identified based on the CoRef resolution process are thus used to expand the knowledge base for the Q-A systems described herein.
In addition, as noted above with respect to the pre-processing functionality of the ingestion engine (e.g., the ingestion engine 126 of
Thus, in some embodiments, the procedure may further include, for at least one segment of the plurality of document segments, identifying at least one segment descriptor comprising one or more of, for example, at least one entity associated with the at least one segment, at least one task associated with at least one segment, and/or subject matter descriptor associated with at least one segment, and tagging the at least one segment with the at least one descriptor. In such example, the procedure may additionally include receiving, at the local device, query data representative of a question relating to source content of the source document, determining at least one query descriptor associated with the query data, the at least one descriptor comprising one or more of, for example, at least one entity associated with the at query data, at least one task associated with the query data, and/or subject matter descriptor associated with the query, and searching a response to the query data from one or more of the plurality of document segments with segment descriptors matching the at least one query descriptor.
Certain implementation can perform analysis of the related documents, and use the results from the related documents to generate appropriate supplemental searches for the main documents managed by the system 100. The agent 110 may also be configured to summarize the related documents accessed, for example, when the content in the related documents is too voluminous to view completely, and users do not know what questions to ask in relation to the related documents. Under those circumstances, a summary of the related documents can be used to suggests what information can be asked. This summary can also be used for automatic identification of question-answer pairs from the content of the related documents.
Thus, in some embodiments, a method is provided that includes receiving at a local device a source document, retrieving augmented data associated with the received source document, and transforming the source document combined with the retrieved augmented data to generate a question-and-answer searchable document.
Embodiments of the method may include one or more of the following features. Retrieving the augmented data may include one or more of, for example, retrieving the augmented data based on links included in the source document, and/or or performing co-reference resolution to determine one or more names or expressions identified in the source document, and searching information repositories to retrieve additional information associated with the determined one or more names or expressions identified in the source document.
Transforming the source document combined with the retrieved augmented data to generate the question-and-answer searchable document may include segmenting the source document combined with the retrieved augmented data into a plurality of document segments.
Transforming the source document combined with the retrieved augmented data may further include transforming each of the plurality of document segments into one or more respective transformed segments according to one or more transformations. Transforming the source document combined with the retrieved augmented data may include transforming the source document combined with the retrieved data according to one or more of, for example, a coarse linearization transform (e.g., according to the fast-match or fast-search procedure) to generate a plurality of vectors representative of content of the source document combined with the retrieved augmented data, and/or a fine-detail transformation to generate high-resolution transformed content records (from which a narrowed range of the location of an answer in the source content can be determined) representative of the content of the source document combined with the retrieved augmented data.
The method may further include deriving a summary of the source document combined with the augmented data based on the transformed question-and-answer searchable document. Deriving the summary may include identifying at least one portion of the transformed question-and-answer searchable document with a level of similarity to a pre-determined summary record, from a repository of summary records, that exceeds a similarity threshold.
As also noted above, the document processing implementations described herein may include different artificial intelligence (AI) techniques to implement variants of vector searching (e.g., implement a practical/productizable transformer-based document Q-A and retrieval). The transformers-based approaches described herein provide high accuracy (better than elastic search), and can be used for automatic Q-A corpus generation (for training), and for “MI” training. Computational effort of such techniques can achieve non-linear scaling (non-linear growth with the size of the document library) that is less than linear scaling. The transformer techniques also offer enhanced customization performance to reduce required training material for large documents, with higher accuracy. Implementations may also include pre-computation of document section vectors, client-specific adaptation of Q-A models, hierarchical fast match, sliding window searching, context augmentation to include more than the literal words (e.g., subject headings) so that limited context of transformer is not too limiting, etc.
In some embodiments, a method is provided that includes receiving one or more source documents associated with a first entity of a plurality of entities, each of the plurality of entities associated with a respective one of question-and-answer models used for processing data, and transforming the one or more source documents associated with the first entity according to the respective associated first question-and-answer model to generate one or more question-and-answer searchable documents.
Embodiments of the method may include one or more of the following features. Transforming the one or more source documents associated with the first entity may include transforming the one or more source documents according to one or more of, for example, a linearization transform to generate a plurality of numerical vectors representative of content of a plurality of document segments for the one or more source documents, and/or a fine-detail transformation to generate fine-detail transformed content records representative of the content of the plurality of document segments. The method may further include applying a linearization transform to one or more of the fine-detail transformed content records to derive resultant fine-resolution numerical vectors for content corresponding to the one or more of the fine-detail transformed content. In some examples, the fine-detail transformation may include a transformation, applied to one or more document segments, based on Bidirectional Encoder Representations from Transformers (BERT) processing. The method may further include receiving query data representative of a question relating to the one or more source documents, and performing a fast search applied to the plurality of numerical vectors generated from application of the linearization transform to segments of the one or more source documents to identify a plurality of match candidates responsive to the query data. The method may further include performing a finer detailed search to identify a match of the query data to portions of content transformed according to the fine-detail transformed objects, with the portions selected according to the plurality of match candidates identified according to the fast search.
In some examples, transforming the one or more source documents associated with the first entity may include transforming the one or more source documents using a neural network trained according to the first question-and-answer model. The method may further include receiving query data representative of a question relating to the one or more source documents, and performing a sliding window search to identify a match of the query data to ordered portions of content transformed according to, for example, a fine-detail transformed objects.
In some embodiments, transforming the one or more source documents associated with the first entity may further include applying one or more pre-processes to the one or more source documents, including segmenting a source document from the one or more source documents into a plurality of document segments. Segmenting the source document may include adding to each segment contextual information representative of the structure and content of the source document. In such embodiments, transforming the one or more source documents may include transforming the segmented source document based, in part, on the contextual information added to each segment.
The method may further include augmenting content of a source document (from the one or more source documents), with the augmenting including one or more of, for example, retrieving the augmented data based on links included in the source document, and/or performing co-reference resolution to determine one or more names or expressions identified in the source document, and searching information repositories accessible from the local device to retrieve additional information associated with the determined one or more names or expressions identified in the source document. Transforming the one or more source documents may include transforming the source document comprising the augmented content. Transforming the one or more source documents may include transforming the one or more source document using, at least in part, pre-computed document segment vectors.
It is to be noted that the various implementations (systems, devices, and methods) described in relation to
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly or conventionally understood. As used herein, the articles “a” and “an” refer to one or to more than one (i.e., to at least one) of the grammatical object of the article. By way of example, “an element” means one element or more than one element. “About” and/or “approximately” as used herein when referring to a measurable value such as an amount, a temporal duration, and the like, encompasses variations of ±20% or ±10%, ±5%, or +0.1% from the specified value, as such variations are appropriate in the context of the systems, devices, circuits, methods, and other implementations described herein. “Substantially” as used herein when referring to a measurable value such as an amount, a temporal duration, a physical attribute (such as frequency), and the like, also encompasses variations of ±20% or ±10%, ±5%, or +0.1% from the specified value, as such variations are appropriate in the context of the systems, devices, circuits, methods, and other implementations described herein.
As used herein, including in the claims, “or” as used in a list of items prefaced by “at least one of” or “one or more of” indicates a disjunctive list such that, for example, a list of “at least one of A, B, or C” means A or B or C or AB or AC or BC or ABC (i.e., A and B and C), or combinations with more than one feature (e.g., AA, AAB, ABBC, etc.). Also, as used herein, unless otherwise stated, a statement that a function or operation is “based on” an item or condition means that the function or operation is based on the stated item or condition and may be based on one or more items and/or conditions in addition to the stated item or condition.
Although particular embodiments have been disclosed herein in detail, this has been done by way of example for purposes of illustration only, and is not intended to be limit the scope of the invention, which is defined by the scope of the appended claims. Any of the features of the disclosed embodiments described herein can be combined with each other, rearranged, etc., within the scope of the invention to produce more embodiments. Some other aspects, advantages, and modifications are considered to be within the scope of the claims provided below. The claims presented are representative of at least some of the embodiments and features disclosed herein. Other unclaimed embodiments and features are also contemplated.
This application claims priority to, and the benefit of, U.S. Provisional Application No. 63/043,906, entitled “Document Processing and Response Generation System” and filed Jun. 25, 2020, the content of which is incorporated herein by reference in its entirety.
| Number | Date | Country | |
|---|---|---|---|
| 63043906 | Jun 2020 | US |
