Patent Application
20230367971
The present disclosure relates to a configuration method of a language resource used for natural language processing by a computer.
In natural language processing by a computer, various data related to a target language prepared in advance are often used. Those data are generally referred to as language resources. There are language resources related to various types of data. Among them, in particular, in language resources related to nouns, the following information is stored.
(1) is attribute data based on a grammatical point of view, such as common nouns, proper nouns, material nouns, and abstract nouns. (2) is data related to classification of the type or concept of the noun, such as a person's name, an organization name, a place name, a date and time, an amount of money, a height, and a distance. (3) is data of knowledge related to relationships present between concepts.
As a representative language resource related to Japanese, there is Japanese meaning outline (see, for example, Non Patent Literature 1). In “GoiTaikei—A Japanese Lexicon”, 300,000 recorded words and 3000 kinds of semantic classifications thereof are defined, and data corresponding to (1) to (3) described above are also recorded for nouns.
In communication, a task of specifying or clarifying content represented by a noun may be required. For example, when content or an entity referred to by a noun appearing in an utterance or text is ambiguous, work of confirming the content or entity is an example of the task.
This task is considered to be composed of two sub-tasks of specifying an entity (specific object or abstract concept) referred to by the noun, and presenting the specified entity. Then, specific processing details required in these sub-tasks varies depending on a type of a target noun or a communication situation or context even for the same noun. Hereinafter, this point will be described in detail.
In the former sub-task, that is, the sub-task of specifying the entity referred to by the noun, what or how much content should be indicated to specify differs depending on the type of the noun, the situation, or the context. For example, in a case where it is necessary to specify a “vehicle of Mr. A”, an issue may be a vehicle type, or an issue may be which vehicle body of a plurality of vehicles parked in front in a parking lot or the like the vehicle is. That is, a noun “vehicle” is a noun in which a vehicle type name may be required to be specified or an individual vehicle body may be required to be specified depending on the situation or the context.
On the other hand, in specifying a noun “vehicle type”, literally the vehicle type name is merely asked, and the individual vehicle body is not required to be specified. In addition, for a noun whose entity is specified as one from the beginning, for example, “Tokyo Tower” or the like, it is not even necessary to perform specification processing in the first place.
In addition, in the latter sub-task described above, that is, the sub-task of presenting the specified entity, what should be or can be indicated and how indication should be or can be performed differ depending on nouns and contexts. For example, in the example of the “vehicle of Mr. A”, in a case where the vehicle type is an issue, the vehicle type name that has been specified may be presented as a voice or a character, that is, a language. However, in a case where specifying the vehicle body is an issue, it is necessary to take a presentation method such as pointing to the vehicle body in front that has been specified, presenting a photograph showing the vehicle body, or presenting a number of a license plate. That is, the noun “vehicle” is a noun for which a method of presenting a specification result differs depending on the situation or the context.
In addition, particularly in natural language processing on a computer, there is a case where a noun specification result is to be presented as a file on the computer in addition to presenting a name and a photograph thereof. For example, if meeting minutes referred to by an utterance “minutes at that time” is edited as a file on the computer, there is also a method of presenting the file itself (with a hyperlink or the like) other than presenting a name of the file also for display of the specification result. That is, the noun “minutes” is also a noun for which the method of presenting the specification result may differ depending on the situation or the context.
As described above, how much it is required to specify the entity or content referred to by the noun or how to present a specification result varies depending on the noun, the situation of communication, or the context.
In a case where a task of specifying content referred to by a noun in communication between humans is performed, humans can appropriately make a detailed determination on the point like the above and select necessary processing.
On the other hand, in natural language processing by a computer, there is no system specialized for a task of specifying the content or entity of a noun in an utterance or text yet. In this regard, not only noun specification processing and presentation processing of a result thereof are not implemented, but also there is no language resource prepared from the viewpoint of execution of such processing. Since the current language resources related to nouns described in the background art do not perform noun classification from the viewpoint as described above and do not assist execution of such a task, under such a current situation, it is not possible to implement a task of specifying the content or entity of a noun in an utterance or text by a computer.
An object of the present disclosure is to provide a method of configuring a language resource related to a noun, which is considered to be necessary for implementing a system that executes a task of specifying content or an entity of a noun in an utterance or text in natural language processing by a computer in consideration of how much it is required to specify the entity or content referred to by the noun or how to present a specification result.
The present disclosure includes a noun classification database that stores nouns in association with, for each noun, information regarding a possible “type of identification operation” and information regarding a “type of presentation method” of an applicable identification result, and searches for, regarding a designated noun, information that specifies or describes an entity of the noun from a background knowledge database on the basis of information regarding corresponding “type of identification operation” and “type of presentation method”.
A data structure of a language resource of the present disclosure is
An utterance understanding support device of the present disclosure
In an utterance understanding support method of the present disclosure,
A program of the present disclosure is a program for implementing a computer as each of functional units included in a communication device according to the present disclosure, and is a program for causing the computer to execute each of steps included in a communication method executed by the communication device according to the present disclosure.
According to the present disclosure, it is possible to provide a method of configuring a language resource related to a noun, which is considered to be necessary for implementing a system that executes a task of specifying content or an entity of a noun in an utterance or text in natural language processing by a computer in consideration of how much it is required to specify the entity or content referred to by the noun or how to present a specification result.
Embodiments of the present disclosure will be described below in detail, with reference to the drawings. Note that the present disclosure is not limited to the following embodiments. These embodiments are merely examples, and the present disclosure can be carried out in a form with various modifications and improvements based on the knowledge of those skilled in the art. Note that components having the same reference signs in the present description and the drawings indicate the same components.
A first embodiment of the present disclosure will be described.
A device of the present disclosure includes a memory that stores language resources of nouns having a data structure of the present disclosure. In the language resources of nouns in the present embodiment, the following classification information regarding a possible “type of identification operation” is held.
In addition, the following classification information regarding a “type of presentation method” of an applicable identification result is held.
Hereinafter, each will be described (note that, in the following, “identification” and “specification” are used in the same meaning, and there is no difference in meaning between them).
In the first place, in noun identification operation,
For example, taking a vehicle as an example, the operation (a) is identification that specifies which vehicle type the vehicle is among cognate objects of vehicles. On the other hand, the operation (b) is identification that specifies a vehicle as an individual (an individual identified by a number of a license plate). The operation (c) is not a description or identification of a target noun itself but operation of specifying another noun corresponding to the noun in terms of content. For example, taking a noun “affiliation” as an example, it is considered that, in specification operation for content referred to by the “affiliation” of a phrase “affiliation of Mr. XX” using the noun, specification of a name of an organization to which “Mr. XX” belongs is required rather than acquiring a description of what the affiliation is. The operation (c) refers to such identification operation.
(1-1) to (1-5) are classifications provided from the viewpoint of which operation among these operations is likely to be requested by each noun in a target language in the identification operation, on the assumption that the pieces of operation such as (a) to (c) exist in the noun identification operation.
In the present embodiment, for each noun of the target language, information (classification tag) on which of these identification operation types the noun belongs to, that is, one of classification tags in a one-to-one correspondence with (1-1) to (1-5) is added. Note that, regarding the “type of identification operation”, one tag is added to one noun due to the nature of the classification method. That is, one noun is not classified into a plurality of items of (1-1) to (1-5).
Hereinafter, these will be described in detail with reference to
First, (1-1) to (1-3) will be described with reference to
For these nouns, “vehicle type” and “family name” in the figure are nouns themselves meaning a type name and a differentiation name of a vehicle or a person, and thus have classification information of (1-1) (in order to indicate this point, a tag “#type name” is also indicated).
In addition, since “Corolla” and “Mr. Tanaka” in the figure themselves are a specific type name and a differentiation name of a vehicle and a person, they serve as results of identification of a type name and a differentiation name of certain nouns. Thus, in a case where these nouns themselves are targets of the identification operation, there is no case other than a case where identification more specialized than the type name and the differentiation name, that is, individual identification is required. Thus, these nouns are nouns having classification information of (1-3) (in order to indicate this point, a tag “#individual identification” is also indicated).
Next, (1-4) will be described with reference to
The same applies to an example of a noun “section manager” in the figure, and in a case of specifying a “section manager” of a phrase “the section manager of the section XX”, it is considered that identification of another noun “person” representing the entity of the “section manager” is required rather than the meaning of the noun “section manager” itself. Note that, for presentation of a result of identification of the noun, a method of presenting a file of a photograph of the person is also conceivable, and thus a tag “#alternative file (#photograph)” for the identification result presentation method is also indicated together.
In addition, nouns such as “support ratio” and “name” in the figure are examples of nouns for which another description is required (thus, a tag “#another description” is also indicated). For example, in a case of specifying a “support ratio” of the phrase “support ratio of XX cabinet”, it is considered that a description of what value the “support ratio” has is required rather than the meaning of the noun “support ratio” itself (thus, a tag “#description presentation (#numeral)” is also indicated). The same applies to the example of the noun “name” in the figure, and in a case of specifying a “name” of a phrase “the name of the section manager of the YY section”, it is considered that a description of what the name is, is required rather than the meaning of the noun “name” itself (thus, the tag “#description presentation (#name)” is also indicated).
Next, examples of a noun having classification information of (1-5) include a noun such as “Tokyo Tower”, that is, a noun that does not require any identification operation because it has a unique entity.
Next, classification information of (2-1) to (2-3) will be described. These are classification information regarding a method of presenting a result of identification of an entity of a noun. In the present embodiment, for each noun of the target language, in addition to the classification tag representing the type of the identification operation described above, classification tags for these types of presentation methods, that is, classification tags in a one-to-one correspondence with (2-1) to (2-3) are added. Note that, for this “type of presentation method”, a plurality of presentation methods may be considered even for the same noun, and thus a plurality of tags may be added to one noun.
In a case where an entity (actual object) of a noun to be identified is a file stored in a computing machine, (2-1) is a noun for which the identification result can be present by presenting the file. Although the entity of the noun to be identified is not a file on the computing machine and the entity file cannot be presented, (2-2) is a noun for which the identification result can be presented by the name of the noun, the result of identification of another noun serving as a description (name or description sentence), a semantically corresponding numeral, or the like. Although the entity of the noun to be identified is not a file on the computing machine and the entity file cannot be presented, there is an alternative file that can present an appearance of the entity of the noun, such as a picture, a photograph, or a symbol, and (2-3) is a noun for which the identification result can be presented by the alternative file.
The above is the description regarding the classification information in the embodiment of the present disclosure. Next, an example in which nouns are classified on the basis of the classification information is illustrated in
An auxiliary tag (#name, #description sentence, #numeral) that defines a type of information used for description is further added to a noun to which the tag (#description presentation) of (2-2) is added. In addition, an auxiliary tag (#picture, #photograph, #symbol) that defines a type of an alternative file is further added to a noun to which the tag (#alternative file) of (2-3) is added. For these nouns, an auxiliary tag to be added is indicated clearly for each noun described in the table.
The main nouns in the table will be described.
It is considered that, for a noun related to a person such as a vehicle, a person, a man, a woman, or the like, the identification result can be presented by description presentation, such as a name (full name), a description sentence that leads to individual identification associated with personal experience, or the like, and the identification result can also be presented by a photograph file obtained by photographing an individual.
It is considered that, for a noun representing a type such as Corolla, Mr. Tanaka, or the like, the identification result can also be presented by a photograph file obtained by photographing an individual, in addition to a description sentence that leads to individual identification associated with personal experience.
For nouns that can themselves be files on a computing machine, such as minutes and materials, it is considered that the identification result can also be presented by names such as titles added to the minutes and materials, in addition to presenting the entity file itself.
It is considered that, for a noun such as a chocolate, the identification result can also be presented by presentation of an alternative file on a computer storing a picture or a photograph of a package, in addition to a description presentation of a name of the noun, a description sentence that leads to individual identification associated with personal experience, or the like.
In the present disclosure, a tag is added assuming that presentation operation for recognizing an individual is executed even for a noun that does not originally need individual identification operation, such as Tokyo Tower. For example, it is operation in a case where a communication party confirms which of several known towers the Tokyo tower is. Thus, as the presentation operation, it is conceivable to present a description sentence that leads to identification associated with personal experience, rather than a general description regarding Tokyo Tower. In addition, it is considered that a file of a picture or a photograph depicting an actual object can be presented.
A second embodiment of the present disclosure will be described.
The present embodiment is an example of a communication system that has data of the language resources described in the first embodiment and uses the content thereof to perform specification of an entity of a noun whose entity in an utterance sentence is ambiguous and perform presentation of a result of the specification.
A configuration of a system of the present embodiment is illustrated in
Each user of the present system participates in communication via the client terminal 30 occupied by the user. The client terminal 30 includes an utterance sentence input unit 31 that inputs utterance of each user and a display screen 32 serving as an interface. The display screen 32 includes an utterance sentence display unit 321 that displays an utterance sentence of each user and a content explanation display unit 322. The utterance sentence display unit 321 has an ambiguous portion designation function for the user to designate a noun appearing therein having an ambiguous entity.
In the server machine 10 different from the client terminal 30, an utterance sentence analysis unit 11, a database search unit 12, and a user interface application 13 operate. The user interface application 13 has a function of receiving an utterance sentence from the utterance sentence input unit 31, analyzing the utterance sentence using the utterance sentence analysis unit 11, searching a background database using the database search unit 12, and controlling the display screen 32, and plays a role of a control module of the entire system.
In addition, on the server machine 10, there is a memory that stores background knowledge data 15 and noun classification data 14. The background knowledge data 15 is data including attribute information of a user of the present system who participates in or is likely to participate in communication, a history regarding communication and various actions, digital information generated as a result thereof, and the like, and is used as a basis in noun entity specification processing described above. The noun classification data 14 is the language resource related to the noun described in the first embodiment of the present disclosure, and is data obtained by digitizing the language resource.
In a case of making an utterance, the user inputs a text sentence having content desired to utter into the utterance sentence input unit 31 of the user's own client terminal 30 illustrated in
The user interface application 13 that has received the text sentence and the identifier of the utterer transmits the received text sentence and the identifier of the utterer to the utterance sentence display units 321 of all the client terminals 30, and adds the information to an utterance history. The user interface application 13 internally accumulates all the utterances of all the users as the utterance history so that a context of an utterance can be grasped.
The utterance sentence display unit 321 of each client terminal 30 has received the text sentence and the identifier of the utterer, and if the received identifier of the utterer is the identifier corresponding to the user of the terminal, displays the received text sentence on the user's own utterance portion of the utterance sentence display unit 321 in
Through the above procedure, communication progresses while the content of the utterance of each user is shared. Having found an ambiguous noun whose entity or content cannot be specified in an utterance sentence by other person or the user himself/herself during progress of communication, the user highlights the portion as in the example of
The user interface application 13 that has received this information uses the utterance sentence analysis unit 11 to execute syntax analysis of a portion designated as the ambiguous portion. Then, a result of the syntax analysis (a noun of the portion designated as the ambiguous portion and information of a modifying portion thereof) is passed to the database search unit 12.
The database search unit 12 that has received the above information searches a table of the background knowledge database unit 15 using the received information, and transmits the acquired inspection result, that is, information serving as the entity or description of the noun expression designated as the ambiguous portion to the user interface application 13. The user interface application 13 forwards the received search result to the content explanation display unit 322 of each client terminal 30.
The content explanation display unit 322 displays the received search result on the display screen 32. As illustrated in
The above is an outline of the operation performed on the display screen of
Operation of the database search unit 12 will be described. The database search unit 12 that has received a search request from the user interface application 13 performs entity specification of a target noun using the noun classification data 14 and the background knowledge data 15 of the present disclosure.
As illustrated in
Thus, the database search unit 12 refers to the noun classification data 14 to inspect whether “#description presentation” or “#alternative file” is added as a tag of the type of presentation method to the noun as the entity specification target (step S1-1 and S1-6), and if each tag is added, further examines a sub-tag (steps S1-2, S1-4, S1-7, S1-9, and S1-11), searches the background knowledge data 15 for content of the sub-tag that seems to represent the entity of the target noun, and sets the content as an entity specification result (steps S1-3, S1-5, S1-8, S1-10, and S1-12).
In this type of noun, since it is not conceivable that “#numeral” is added as a sub-tag of “#description presentation”, the database search unit 12 does not inspect the presence or absence of this sub-tag. Note that, in the present disclosure, the content and format of the background knowledge data 15 and a specific method of specifying the entity of the target noun using the background knowledge data 15 are not defined.
When it is determined that the individual identification is also required, the database search unit 12 executes the processing of steps S2-2 to S2-13. That is, as in the case of
In a case where it is determined that the individual identification is not required, the database search unit 12 executes the processing of steps S2-14 to S2-25 of
As described above, the present disclosure is characterized by constructing, for each noun, a language resource that holds information regarding a possible “type of identification operation” and a “type of presentation method” of an applicable identification result.
Specifically, the following classification information is held for the possible “type of identification operation”.
In addition, the following classification information is held for the “type of presentation method” of the applicable identification result.
The present disclosure constructs, for each noun, a language resource that holds information regarding a possible “type of identification operation” and a “type of presentation method” of an applicable identification result, and thus, it is possible to solve the problem of the present disclosure, that is, a problem that there is no language resource related to a noun necessary for implementing a system that executes a task of specifying content or an entity of a noun in utterance or text, in consideration of how much it is required to specify the entity or content referred to by the noun or how to present a specification result.
The present disclosure can be applied to the information communication industry.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/JP2020/034745 | 9/14/2020 | WO |
