Patent Application
20230146105
Embodiments of the present invention relate to a technique for assisting creation of knowledge information such as FAQs.
Techniques for assisting interactions by means of FAQs have conventionally been used. FAQs or frequently asked questions are combinations of questions and answers created, for example, from user support histories. FAQs are used by operators serving users in contact centers or are put on websites as a guidance tool for users to find answers of questions or solutions of problems.
It is an object of the present invention to provide a knowledge information creation assist apparatus allowing knowledge information such as FAQs to be created from user support histories accumulated in the past with a reduced workload.
According to an embodiment, a knowledge information creation assist apparatus provides a function of assisting creation of knowledge information related to questioning. The knowledge information creation assist apparatus includes: an input control section configured to control upload of a plurality of user support histories including questions; a question utterance extraction section configured to perform processing of removing at least one unnecessary utterance sentence not related to questioning from the uploaded user support histories or processing of extracting at least one utterance sentence related to questioning from the user support histories and then output the resulting user support histories as a result of the processing; a grouping control section configured to classify the user support histories output as the result of the processing from the question utterance extraction section into a plurality of question groups according to a predetermined classification model; and a screen control section configured to perform first display control for displaying, on a display apparatus, a question group creation screen including group information of each of the question groups and second display control for displaying, on the display apparatus, a knowledge information creation screen for producing knowledge information in response to input performed by a creator to one of the question groups selected on the question group creation screen.
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
[
An embodiment of the present invention will be described with reference to the accompanying drawings.
An example of knowledge information is FAQs (Frequently Asked Questions), and the FAQs are formed of combinations of questions and answers (the combinations of questions and answers do not necessarily have one-to-one correspondence but may have one-to-n or n-to-one correspondence). As described above, the FAQs can be used by operators serving users in contact centers to achieve efficient responses to users, thereby streamlining the operation in contact centers. The FAQs can also be put on websites to help users solve their questions or problems with reference to the FAQs, which can enhance convenience of users and can reduce the number of inquiries (the number of incoming calls) in contact centers.
In general, the FAQs are often created by managers based on a large amount of user support logs accumulated in the past (user support histories), and involves a significant burden and cost in creation and maintenance (update). Since all the large amount of user support logs cannot be reviewed, the quality of the FAQs may depend on the managers.
To reduce the burden of creation and maintenance of the FAQs, techniques have been proposed for automatically classifying user support logs to automatically create optimal FAQs (For example, see Patent Documents 1, 2). However, the automatically created FAQs cannot necessarily be put to practical use as they are. In reality, those FAQs are corrected by managers, or in some cases, FAQs end up being created based on manual operation.
To address this, the assist apparatus 100 according to Embodiment 1 provides an environment in which FAQs can be created manually from a large amount of user support logs without cumbersome operation. Specifically, the assist apparatus 100 provides the following principal functions of assisting creation of knowledge information through a user interface function:
The assist apparatus also provides (A) a function of extracting question utterances (function of filtering question utterances) performed before the grouping (2) of the user support logs. The details of the question utterance extraction function will be described later.
First, description is made of an example of accumulation of user support logs and use of FAQs in conjunction with the example shown in
In a contact center system, an operator interacts with a user via telephone (call), chat, or e-mail to answer a question (inquiry) from the user. A user support log in which the interaction is recorded is managed and accumulated in a customer support management apparatus 330. The customer support management apparatus 330 can be used to provide a function of assisting the operator in user support and perform processing of providing FAQs through an FAQ search function.
The customer support management apparatus 330 can accumulate the user support log, for example, in an operator input mode in which a predetermined input screen is displayed on an operator terminal 300 to allow the operator to input the user support log (mainly the content of the question and the answer).
Alternatively, the content of voice interaction with the user via a user terminal 500 can be converted into text through voice recognition and the text can be compared with existing FAQs to retrieve and display any FAQ appropriate for the content of the interaction on a display apparatus of the operator terminal 300. The same applies to chat, in which case the content of interaction through time-series text messages can be used to retrieve and display any existing FAQ appropriate for that content. In this case, the customer support management apparatus 330 can automatically manage and accumulate the series of interactions in text form as a user support log (automatic input mode).
In addition to the automatic search mode involving automatic matching with the content of interaction, the FAQ search function includes, for example, a manual search mode in which the operator inputs a keyword to perform manual search on an FAQ search screen and then any suitable existing FAQ is displayed as the result of the search. In the manual search mode, the operator can also search a list of existing FAQs to retrieve and display any suitable FAQ by himself. The FAQ search function can allow both of the automatic search mode and the manual search mode.
In the automatic search mode described above, the matching between the content of interaction and the FAQs can be performed by means of a query model and a search model using a classification learner. The classification learner adopts various techniques, and for example, can quantify (weight) individual words in a document based on the number of appearances or the similarity between the words to utilize a known technique such as a Naive Bayes classifier which is one of classification techniques in machine learning. The classification technique is not limited to the Naive Bayes classifier, and other known techniques may be used.
The query model is, for example, a classification learner which is trained to extract a sentence intended to question from a group of sentences. The search model is a classification learner which is trained to extract an FAQ (answer) appropriate for the sentence (question) extracted in the query model. Since the techniques for grouping and sorting sentences using the classification learner (model) are known technologies, detailed description thereof is omitted.
While the following description is made of FAQ creation assist based on the user support logs accumulated in the single contact center, the present invention is applicable in an aspect in which contact centers are deployed in a plurality of sites. For example, the assist apparatus 100 can be connected to each of the contact centers (customer support management apparatuses 330) in those sites over a network and acquire user support logs from the contact centers to create FAQs. In this case, the assist apparatus 100 can create an integrated FAQ from the plurality of sites or create a different FAQ for each site.
In this manner, the customer support management apparatus 330 accumulates user support logs on a daily basis and provides user support using FAQs. A manager can extract and upload a plurality of user support logs from the customer support management apparatus 330 to the assist apparatus 100 according to Embodiment 1 to create an FAQ through the use of the knowledge information creation assist function. The created FAQ can be output in a predetermined data format (for example, a CSV format). The output FAQ information can be input to the customer support management apparatus 330 and used for user support based on FAQs (including putting the FAQ information on a website).
The control apparatus 120 includes a user support log input control section 121, a grouping control section 122, a screen control section 123, an existing FAQ association control section 124, a knowledge information creation section 125, an output control section 126, and a question utterance extraction section 122A.
The user support logs are formed of combinations of questions and answers and are included in an utterance history from interactions between users and operators, a user support history in the operator input mode, a chat interaction history and the like. It should be noted that questions and answers may not be unambiguously determined such that the question is what a user says and the answer is what an operator says. Thus, the “question” may include what an operator says when he repeats a user’s question or says it in other words.
The utterance history includes text data created from voice interactions between users and operators through voice recognition processing, and can be accumulated separately in user’s utterances and operator’s utterances as shown in
These user support logs are created by the customer support management apparatus 330. The voice recognition processing can be performed by a voice recognition processing section incorporated in the customer support management apparatus 330 or an external voice recognition processing server.
A user support log file uploaded to the assist apparatus 100 may correspond to data (file) from a single user support interaction or from a different one of operators, or may be file data including a plurality of user support logs provided by collecting all user supports. A plurality of questions may be included in each of the utterance history, the user support history, and the chat interaction history. For example, in an interaction between a user and an operator, the user may ask questions in succession over time, and the operator may successively answer the questions. As a result, each of the histories may include a plurality of user support logs (a plurality of questions and answers).
The question utterance extraction section 122A performs question utterance filtering processing on the acquired user support log file. The question utterance filtering processing includes removing any utterance corresponding to a filler (an utterance element which is meaningless in itself but is used to fill a gap between utterances or to probe or get the hint of the reaction of an interaction partner) or a greeting included in the user support file to extract a group of utterances consisting mainly of question sentences. The grouping processing is performed on the result of the question utterance filtering processing.
Next, the function (2) of grouping and displaying the user support logs includes, with a predetermined classification leaner, grouping “questions” found in the uploaded user support logs which have been subjected to the question utterance filtering processing (A). The grouping is performed in the manner described above. For example, words in question sentences can be quantified (weighted) based on the number of appearances or the similarity between the words to group similar question sentences or sentences having similar intention. The function (2) is performed by the grouping control section 122.
The grouping control section 122 can automatically extract a representative sentence (representative question sentence) and keyword(s) representative of each of the groups. The extracted representative sentence can be used as a group name.
The representative sentence and keyword(s) can be produced, for example, by scoring the number of appearances of each word or the similarity between words within the question sentences. Specifically, a word having a high number of appearances in the grouped question sentences can be extracted as a keyword, or the grouped question sentences can be evaluated with a query model to select (extract) a question sentence having a high evaluation level as a representative sentence.
The result of the grouping is displayed on a screen as group information. Existing FAQs can be displayed together with the result of the grouping. The function (3) of reading and displaying existing FAQs includes displaying already created FAQs (already created knowledge information) created previously (in the past) with the knowledge information creation assist function and/or created previously without the knowledge information creation assist function. The function (3) is performed by the existing FAQ association control section 124 which reads the existing FAQs previously stored in the storage apparatus 130 and displays the read FAQs on the screen together with the result of the grouping. As described later, the existing FAQ association control section 124 can also perform processing of association display in response to operation on the screen such as emphasizing an existing FAQ similar to any of the questions in the groups.
The function (4) of creating an FAQ based on the result of the grouping includes selecting group information (for example, group display panel) resulting from the grouping and creating an FAQ from the group information. In response to selection of group information, an FAQ creation screen (knowledge information creation screen) is displayed. On the FAQ creation screen, an FAQ can be created based on input operation performed by the manager (operator).
On the FAQ creation screen, an FAQ edit screen block and a reference information display block are displayed side by side. With reference to information in the reference information display block, the manager performs input and edit in the FAQ edit screen block. The manager can select a “determine” button to create (produce) one FAQ belonging to one group. The created FAQ is assigned predetermined FAQ identification information (FAQ_No) and stored in the storage apparatus 130. The FAQ edit and creation processing is performed by the knowledge information creation section 125.
The function (5) of outputting the created FAQ includes outputting, in the CSV format, the FAQ assigned unique FAQ identification information stored in the storage apparatus 130. In response to selection of a predetermined “output as FAQ” button, the output control section 126 converts the FAQ into an output file of the CSV format and transmits the file to the manager terminal 350.
The output control section 126 can perform control to output only the newly created FAQ, that is, the difference from the existing FAQs, or to output all the FAQs including the newly created FAQ as an existing FAQ. Any output range of FAQs may also be set. The output format is not limited to the CSV format, and for example, any data format may be used as long as the customer support management apparatus 330 can acquire the data.
The screen control section 123 controls the overall display on the screens where the functions (1) to (5) are provided. The display control for the individual functions can be performed by the associated functional sections, and those functional sections cooperate with the screen control section 123 to control the overall display processing.
First, the manager connects to the customer support management apparatus 330 from the manager terminal 350 and extracts user support logs including question and answer sentences in the form of text data. The customer support management apparatus 330 can have a function of outputting the user support logs in the CSV format, for example. A user support log file corresponds to data including a plurality of user support logs.
As shown in
The user support log data display block includes a “create question group” button. The manager can select the “create question group” button (S303) to create a question group from any user support log file (a plurality of user support logs) displayed within the list in the user support log data display block.
In Embodiment 1, the question utterance filtering processing is performed on the user support log file to be processed by the grouping control section 122 to remove any utterance corresponding to a filler or a greeting included in the user support log file (S103A). Then, the grouping control section 122 performs grouping processing on the user support log file including a group of utterances consisting mainly of question sentences from which the unnecessary utterance phrases and sentences have been removed.
The grouping control section 122 performs question grouping processing on the one or more uploaded user support log files which have been subjected to the question utterance filtering processing by grouping the questions included in the user support logs through the use of the classification learner described above (S103). The screen control section 123 displays the resulting question group in the question group creation display block.
The grouping control section 122 counts the user support logs included in one question group. The grouping control section 122 performs processing of extracting a representative sentence and keyword(s). The question group name (or group ID), the number of belonging user support logs, and the representative sentence and keyword(s) are stored in the storage apparatus 130 as grouping related information for each question group. The grouping control section 122 creates information indicating the correspondence between each question group and the belonging user support logs and stores the information in the storage apparatus 130. For example, each question group can be associated with the IDs of the user support logs, or each group name (group ID) can be associated with the user support logs.
After the grouping processing, the screen control section 123 creates question group display information including at least the representative sentence, the number of belonging user support logs, and the keyword(s), and displays the created information in the question group creation display block (S104). An example of the question group display information is group display panels in the form of rectangular icons as shown in
The question group creation display block is now described. In the question group creation display block, the manager can select one of display styles including list and panel styles and display the question groups in the selected style (S304). The example of
The number of question groups and a cover ratio are displayed in the question group creation display block. The cover ratio refers to the ratio of the total number of the user support logs belonging to the groups displayed in the question group creation display block to the total number of the user support logs uploaded to the user support log data display block. In the example of
With the cover ratio calculated and displayed in this manner, it is possible to know how many user support logs can be covered by the question groups which have been narrowed down, or in other words, to know how many question groups are needed to cover all the user support logs. Specifically, as the number of question groups is increased, the number of created FAQs is increased, and an increased number of FAQs allows detailed support tailored to the individual needs of users. However, an extremely increased number of FAQs provides too many options in user support and is likely to offer FAQs which do not exactly fit the needs of users. Such an increased number of FAQs also puts a great burden on the operator searching FAQs. Thus, instead of simply creating an FAQ from each of all the question groups covering 100% of the user support logs, creating FAQs from a set of question groups which are low in number but high in cover ratio can achieve FAQs which are efficient and likely to fit the needs of users. In addition to the viewpoint of the number of user support logs, the cover ratio can be calculated while taking account of information about time taken for user support and operator information acquired from the customer support management apparatus 330. As a result, FAQs can be created from questions requiring longer times for answers by giving a higher priority to those questions, or FAQs can be created from user support logs including a highly regarded (experienced) operator by giving a higher priority to those user support logs.
In the example of
The existing FAQ display block includes a “read existing FAQ” button. When the manager selects the “read existing FAQ” button, the existing FAQs stored in the storage apparatus 130 can be displayed (S305).
As shown in
The “divide” button can be used to perform the question grouping processing again on user support logs included in a specified one of the question groups to create two or more question groups and stores them in the storage apparatus 130. The divided question groups are again subjected to the processing of extracting a representative sentence and keyword(s).
Next, the manager can select a question group to create an FAQ. When a group display panel is selected (S306), the screen control section 123 displays an FAQ creation screen shown in
The screen control section 123 controls input to each input field and also controls selection of selection items including a result of matching with existing FAQs, category, and bookmark (priority). The selection items are associated with the narrowing conditions described above, and are items manually setting by the manager.
A reference information display block on the FAQ creation screen has a plurality of tabs. The manager can select one of the tabs to display reference information linked thereto (S307). The example of
In the “associated user support history,” the user support logs belonging to the question group are displayed. The screen control section 123 can extract and display the belonging user support logs based on the correspondence between question groups and user support logs stored in the storage apparatus 130 (YES at S108, S109). As shown in
With reference to various types of information in the reference information display block on the FAQ creation screen, the manager can input and edit a question sentence, input an answer sentence, and perform various selection operations of the selection items (S114, S115).
To create an FAQ from the combination of the input question sentence and answer sentence, the manager selects the “determine button” in the FAQ edit screen block (S308). In response to the selection of the “determine” button, the knowledge information creation section 125 assigns FAQ identification information (FAQ_No) to the set of the question sentence in the question sentence input field and the answer sentence in the answer sentence input field to create an FAQ (S116). The knowledge information creation section 125 stores the created FAQ in the storage apparatus 130 (S117).
After the creation of the FAQ on the FAQ creation screen, the screen control section 123 performs control to transition from the FAQ creation screen to the question group creation screen, thereby displaying the question group creation screen on the manager terminal 350. The screen control section 123 controls the display of the question group associated with the created FAQ such that the question group has the status updated to “created” (S118). For example, in the example of
After the creation of the FAQ on the question group creation screen and the FAQ creation screen, the manager can output the FAQ for use in contact centers. In response to selection of the “output as FAQ” button” on the question group creation screen, the output control section 126 creates, for example, FAQ output data in the CSV format from the FAQ having the “created” status in the question groups displayed on the question group creation screen (S119). The output control section 126 can perform processing of updating the status of the FAQs stored in the storage apparatus 130 by setting the FAQ to be output as an existing FAQ (S120). With this configuration, the FAQ from the question group displayed on the present question group creation screen can be controlled not to be displayed as an existing FAQ and can be controlled to be displayed as an existing FAQ in the next creation processing.
The output control section 123 transmits the created FAQ output data to the manager terminal 350 (S121). The manager stores the FAQ downloaded from the assist apparatus 100 in the manager terminal 350 or the external storage apparatus. The manager then applies (acquires) the created FAQ to the customer support management apparatus 330 and utilizes the FAQ for support operations in the contact center.
As in the example of
To identify the information related to questioning in the user support log file, an utterance classifying dictionary is used to distinguish utterances (non-question utterances) such as fillers and greetings having no information representing questioning from utterances having information representing questioning. One or more utterance sentences having information representing questioning are extracted and input to the grouping control section 122.
The utterance classifying dictionary includes a first library including accumulated words, keywords, and groups of exemplary utterances having no information representing questioning and a second library including accumulated words, keywords, and groups of exemplary utterances having information representing questioning. Each of the libraries can be formed as a customer utterance library or an operator utterance library. For example, customized libraries can be provided such that the characteristics of individual customers are taken into account by accumulating examples of utterances having no information representing questioning and examples of utterances often having information representing questioning which are frequently found in inquiries from the customers. Utterance classifying dictionaries can be created and updated for different business types, products, regions, or participants in interactions (customer or operator) using a well-known machine learning function. In addition to humans such as the operator, the computer-based interactive voice response (IVR) system can be regarded as an utterer in the user support history. In other words, the user support history including utterance sentences provided from the IVR system and users can be covered by the scope in which Embodiment 1 is implemented.
Specifically, as shown in
This modular configuration can ensure independence between the modules and association between the modules to customize the question utterance extraction function with ease and at low cost appropriately for different business types, products, regions, and participants in interactions.
Referring back to
In the example of
After the specification of the extraction start position, the question utterance extraction section 122A extracts a predetermined number of utterances subsequent to the start position as a question sentence unit (processing (b) in
It should be noted that the predetermined number N may be set to any number. For example, the history of customer-operator interactions can be previously analyzed through statistical processing, and the result of the analysis can be used to determine the range to be extracted as the question sentence unit.
Another implementation includes checking the utterances sequentially from the start position to see whether or not a particular greeting uttered by the operator is present and determining the particular greeting to be the end position, instead of using the predetermined number N. For example, the operator may say his name at the end of an interaction as “I’m OO in charge of customer support. Thank you for your continued support.” Such a sentence suggesting the end of the user support can be regarded as a special sentence which can be determined to be the end position, thereby defining the range to be extracted as the question sentence unit.
The question utterance extraction section 122A creates a user support log which consists of the extracted question sentence unit, that is, from which the utterance sentences having no information representing questioning have been removed, and outputs the created user support log to the grouping control section 122 (processing (c) in
It should be noted that utterance sentences in a user support log file of one customer may be segmented at predetermined time intervals and the question utterance extraction processing may be performed for each segment. In this case, the file may be divided into sub-files at predetermined time intervals and the processing may be performed for each sub-file, or the processing may be performed on the respective segments of the file in stages without dividing the file.
Conversely, a plurality of user support logs may be combined into one which in turn may be subjected to the question utterance extraction processing. In this case, user support logs may be combined into one according to conditions such as logs from the same customer, logs from the same operator, or logs during a predetermined time period being combined into one user support log, and then the processing may be performed on the one user support log.
The knowledge information creation assist system according to Embodiment 1 can create question groups from a large amount of user support logs to relieve the workload in creating FAQs corresponding to knowledge information. If too many question groups are created, it is troublesome to select an appropriate one from them. Especially, if many utterance sentence groups less related to questioning are present, an appropriate question group may be buried in those many question groups.
In the knowledge information creation assist system according to Embodiment 1, the question groups are created and displayed on their respective panels in the question group creation display block on the question group creation screen. Since the display area on the screen is limited, the user should scroll on the screen to see the many question groups. Considering the possibility of missing an appropriate question group, it is desirable to avoid display or creation of question groups less related to FAQ creation as much as possible.
In Embodiment 1, the question utterance extraction function can remove unnecessary utterance sentences to result in a reduction in number of question groups to be created in the grouping processing performed by the grouping control section 122.
For example, in the example of
It should be noted that the extraction start position in the utterance sentences within the user support log file can be determined in response to an utterance of an operator, in addition to an utterance of a customer. An example of such an utterance of an operator is a question utterance sentence intended to promote a user to make a replay. For example, the operator says “You saw a commercial, so are you inquiring about the cheap insurance premiums of people in their fifties?” and the customer replies “Yes.” In this case, the phrase “are you inquiring about” can be registered as a set phrase in the utterance classifying dictionary and this set phrase can be fixedly specified (determined) to be the extraction start position.
The question utterance extraction function according to Embodiment 1 is summarized as follows.
The knowledge information creation assist system according to Embodiment 1 allows manual creation of FAQs from a large amount of user support logs without cumbersome operation to lower the cost in creating and maintaining the FAQs. Specifically, the question utterance extraction function according to Embodiment 1 can remove the unnecessary utterance sentences to reduce the number of question groups to be created. This can reduce the number of the displayed question groups in the display block to prevent any question group having information representing questioning from being buried in the question groups having no information representing questioning, thereby improving the operation efficiency in creating the FAQs.
While the embodiment has been described, the knowledge information creation assist apparatus 100 and the customer support management apparatus 330 may be implemented as an integrated computer system. In other words, the customer support management apparatus 330 may be configured to include the functions of the knowledge information creation assist described above.
The user terminal 500 includes a multi-functional cellular phone such as a smartphone, a mobile communication terminal apparatus such as a Personal Digital Assistant (PDA), and an information processing terminal apparatus such as a personal computer having a calling function, a communication function, and a computing function. Each of the operator terminal 300 and the manager terminal 350 is implemented by a computer apparatus equipped with a browser which performs processing of displaying the various screens described above.
The knowledge information creation assist apparatus 100 can include, as a hardware configuration, a memory (main storage apparatus), operation input means such as a mouse, keyboard, touch panel, and scanner, output means such as a printer, and an auxiliary storage apparatus (such as a hard disk), in addition to the components described above.
The functions of the present invention can be implemented by a program. A computer program previously provided for implementing each function can be stored on an auxiliary storage apparatus, the program stored on the auxiliary storage apparatus can be read by a control section such as a CPU to a main storage apparatus, and the program read to the main storage apparatus can be executed by the control section to allow a computer to perform the function of each component in the present invention. Each of the functions of the present invention can also be implemented by a different one of apparatuses, and those apparatuses can be connected directly or via a network to constitute a computer system.
The program may be recorded on a computer readable recording medium and provided for the computer. Examples of the computer readable recording medium include optical disks such as a CD-ROM, phase-change optical disks such as a DVD-ROM, magneto-optical disks such as a Magnet-Optical (MO) disk and Mini Disk (MD) , magnetic disks such as a floppy disk® and removable hard disk, and memory cards such as a compact flash®, smart media, SD memory card, and memory stick. Hardware apparatuses such as an integrated circuit (such as an IC chip) designed and configured specifically for the purpose of the present invention are included in the recording medium.
While the embodiment of the present invention has been described, the embodiment is only illustrative and is not intended to limit the scope of the present invention. The novel embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made thereto without departing from the spirit or scope of the present invention. The embodiment and its variations are encompassed within the spirit or scope of the present invention and within the invention set forth in the claims and the equivalents thereof.
100
110
120
121
122
122A
123
124
125
126
130
300
330
350
500
| Number | Date | Country | Kind |
|---|---|---|---|
| 2020-057551 | Mar 2020 | JP | national |
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2021/010479 | 3/16/2021 | WO |
