JP7074940B1 - Entity identification device and entity identification method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an entity identification device and an entity identification method capable of appropriately identifying a person appearing in a text document. SOLUTION: An entity identification device identifies a person name corresponding to the first term by using an extraction unit for extracting a first term related to a person appearing in a text document and a database storing a name related to the person. It is provided with an identification unit for identifying a target person indicated by the first term based on the specified person designation. [Selection diagram] Fig. 2

Description

The present invention relates to an entity identification device and an entity identification method.

Conventionally, communication tools such as chat and e-mail have been introduced in organizations such as companies. With such a communication tool, the amount of text documents such as messages becomes enormous.

In addition, by analyzing text documents, it is assumed that they will be used for classification of text documents. For example, a technique has been proposed in which the meaning of a text document is changed by replacing, adding, or deleting words contained in the text document, and the text document is classified hierarchically by using machine learning (for example, Patent Document 1).

Japanese Unexamined Patent Publication No. 2018-106390

Against the background described above, the inventors have found that it is extremely important to identify the person who appears in the text document for the communication tool used in the organization as a result of diligent studies. For example, by identifying a person who appears in a text document, a text document related to that person can be extracted, and business can be facilitated.

However, in communication tools, it is rare that a person's name (full name) is used in a text document, and only the person's surname is used, or a person's synonym (nickname, etc.) is often used.

On the other hand, in the above-mentioned machine learning, teacher data is required and it is not specialized in identifying a person, so that the generation of teacher data is complicated and it is not possible to identify a person appropriately.

Therefore, the present invention has been made to solve the above-mentioned problems, and an object of the present invention is to provide an entity identification device and an entity identification method capable of appropriately identifying a person appearing in a text document. do.

One aspect of the disclosure is to identify and identify the person name corresponding to the first term by using an extraction unit for extracting the first term related to the person appearing in the text document and a database storing the name related to the person. It is an entity identification device including an identification unit for identifying a target person indicated by the first term based on a person designation.

One aspect of the disclosure is specified as a step of extracting a first term relating to a person appearing in a text document and a step of identifying a person designation corresponding to the first term using a database storing names relating to the person. It is an entity identification method including a step of identifying a target person indicated by the first term based on a person name.

INDUSTRIAL APPLICABILITY According to the present invention, it is possible to provide an entity identification device and an entity identification method that can appropriately identify a person appearing in a text document.

FIG. 1 is a diagram showing an entity identification system 100 according to an embodiment. FIG. 2 is a diagram showing an entity identification device 30 according to an embodiment. FIG. 3 is a diagram showing an example of data stored in the management unit 32 according to the embodiment. FIG. 4 is a diagram for explaining a specific example according to the embodiment. FIG. 5 is a diagram for explaining a specific example according to the embodiment. FIG. 6 is a diagram for explaining a specific example according to the embodiment. FIG. 7 is a diagram for explaining a specific example according to the embodiment. FIG. 8 is a diagram for explaining a specific example according to the embodiment. FIG. 9 is a diagram for explaining a score according to an embodiment. FIG. 10 is a diagram showing an entity identification method according to an embodiment.

Hereinafter, embodiments will be described with reference to the drawings. In the description of the drawings below, the same or similar parts are designated by the same or similar reference numerals.

However, it should be noted that the drawings are schematic and the ratio of each dimension may differ from the actual one. Therefore, the specific dimensions should be determined in consideration of the following explanation. In addition, it goes without saying that there may be a portion where the relations or ratios of the dimensions of the drawings are different from each other.

[Summary of disclosure]
The entity identification device according to the outline of the disclosure identifies the person name corresponding to the first term by using the extraction unit that extracts the first term related to the person appearing in the text document and the database that stores the name related to the person. , The identification unit for identifying the target person indicated by the first term based on the specified person designation.

The entity identification method according to the outline of the disclosure includes a step of extracting the first term related to a person appearing in a text document and a step of identifying a person name corresponding to the first term using a database storing the name related to the person. And a step of identifying the target person indicated by the first term based on the identified person designation.

According to the summary of the disclosure, the entity identification device extracts the first term relating to the person appearing in the text document, identifies the person designation corresponding to the first term, and based on the identified person designation, the first term. Identify the target person indicated by. According to such a configuration, the target person (entity) that appears in the text document can be appropriately identified.

Although not particularly limited, the text document may be the content of a communication tool such as chat or email. Therefore, the sender of the text document may be known. Similarly, the addressee of the text document may be known. The sender may be read as a writer, a speaker, or the like. Under such an assumption, the target person identified by the entity identification device may be a person other than the sender and the addressee.

[Embodiment]
(Entity identification system)
Hereinafter, the entity identification system according to the embodiment will be described. FIG. 1 is a diagram showing an entity identification system 100 according to an embodiment. Although not particularly limited, the entity identification system 100 identifies the target person indicated by the first term appearing in the text document. The text document may be the content of a communication tool such as chat or email.

Further, the communication tool may be a tool used in the organization. The organization may be an organization having legal personality or an organization without legal personality. The person may be a person who has an employment relationship with the organization or a person who has a contractual relationship with the organization. Although not particularly limited, the following exemplifies a case where the organization is a corporation and the person is an employee.

As shown in FIG. 1, the entity identification system 100 includes a first terminal 10, a second terminal 20, and an entity identification device 30. The first terminal 10, the second terminal 20, and the entity identification device 30 are connected by the network 200. Although not particularly limited, the network 200 may be configured by an Internet network. The network 200 may include a local area network, a mobile communication network, or a VPN (Virtual Private Network).

The first terminal 10 is a terminal used by the sender of the text document. For example, the first terminal 10 may be a personal computer, a smartphone, or a tablet terminal. Although not particularly limited, the sender may input a text document using the first terminal 10.

The second terminal 20 is a terminal used by the addressee of the text document. The second terminal 20 may be a personal computer, a smartphone, or a tablet terminal. Although not particularly limited, the addressee may view the text document using the second terminal 20.

Although not particularly limited, the addressee may be two or more addressees. For example, when the communication tool is e-mail, not only the addressee specified by TO but also the addressee specified by CC (Carbon Copy) or BCC (Blind Carbon Copy) may be included. When the communication tool is a group chat, the addressee may be a member included in the group chat. Alternatively, when the function specified by the addressee (mention function) is used, the addressee may be a member designated by the mention function.

The entity identification device 30 is a device for identifying a target person indicated by the first term appearing in a text document. The details of the entity identification device 30 will be described later.

(Entity identification device)
Hereinafter, the entity identification device according to the embodiment will be described. FIG. 2 is a diagram showing an entity identification device 30 according to an embodiment. As shown in FIG. 2, the entity identification device 30 includes a communication unit 31, a management unit 32, and a control unit 33.

The communication unit 31 is composed of a communication module. The communication module may be a wireless communication module compliant with standards such as IEEE802.11a / b / g / n / ac / ax, LTE, 5G, and 6G, and a wired communication module compliant with standards such as IEEE802.3. May be.

In the embodiment, the communication unit 31 receives the text document. For example, the communication unit 31 may receive a text document from a database that stores a text document transmitted or received using a communication tool. Alternatively, the communication unit 31 may cooperate with the communication tool and receive the text document at the stage when the text document is transmitted or received by using the communication tool.

The management unit 32 is composed of storage media such as SSD (Solid State Drive) and HDD (Hard Disk Drive), and stores various information.

In the embodiment, the management unit 32 may have a database (hereinafter referred to as a first DB) for storing names related to a person. The management unit 32 may have a database (hereinafter referred to as a second DB) for storing names related to the affiliation of a person. The management unit 32 may have a database (hereinafter referred to as a third DB) for storing the correlation related to the correlation of people. The management unit 32 may have a database (hereinafter, 4th DB) for storing characteristic words related to a person. Although not particularly limited, the first DB, the second DB, the third DB, and the fourth DB may be included in the personnel DB that stores data related to the personnel of employees belonging to the corporation.

For example, the management unit 32 may have a personnel database for storing the data shown in FIG. As shown in FIG. 3, the management unit 32 may manage data associated with the first designation, the second designation, the affiliation designation, the correlation, and the characteristic word.

The first designation is a designation that includes at least a part of the name and name. The first name may be referred to as a real name, a real name, or the like. In the embodiment, only "Mr." may be considered as the first designation, and only "name" may be considered as the first designation. The first designation is an example of a person designation.

The second designation is a designation other than the designation including at least a part of the name and the name. The second name may be referred to as a nickname, another name, a synonym, another name, or the like. The second designation is associated with at least the first designation. The second designation is an example of a person designation.

The affiliation name is a name related to the affiliation of the person who has the first name. The affiliation designation is associated with at least the first designation. Although not particularly limited, the affiliation designation may be hierarchically associated with the first designation. For example, the affiliation name for each layer such as XX headquarters, XX department, XX section, and XX charge may be associated with the first name.

The correlation is information that identifies a person who has a correlation with a person who has the first designation. The correlation is associated with at least the first designation. The correlation includes information representing the correlation distance to the person having the first designation. For example, the correlation distance to a person who belongs to the same department as the person having the first designation may be short, and the correlation distance to the person belonging to a different department from the person having the first designation may be long. The correlation distance between the person having the first designation and the person having a business connection may be short, and the correlation distance between the person having the first designation and the person having no business connection may be long. The correlation distance between the person who has the first designation and the person who is personally close may be close, and the correlation distance between the person who has the first designation and the person who is not personally close may be long. The correlation may be specified based on a correlation diagram (for example, an organization chart) of the departments constituting the organization, may be input by a person having the first designation, or may be input by a third party such as a personnel person in charge. It may be done.

The characteristic word is a characteristic word relating to a person having the first designation. The feature word is associated with at least the first designation. The characteristic term may include a term that expresses the business-related characteristics of the person having the first designation, or may include a term that expresses the personal characteristics of the person having the first designation. Feature words may include abbreviations for these terms.

Figure 3 illustrates a case where he has three employees in common with "AA". The person whose first name is "AA XX" belongs to the marketing department, and the person whose first name is "AA YY" belongs to the legal office, and the first name is "AA ZZ". The person who is belongs to the sales department.

The control unit 33 may include at least one processor. At least one processor may be composed of a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), one or more Integrated Circuits, one or more Discrete Circuits, and a combination thereof. ..

In the embodiment, the control unit 33 has an extraction unit 33A and an identification unit 33B.

The extraction unit 33A constitutes an extraction unit that extracts the first term related to the person appearing in the text document. Since the first term is a term that appears in a text document, it may be the first name (name only, name only, name), or it may be the second name. Although not particularly limited, the first term may be extracted by a known morphological analysis or the like.

The identification unit 33B identifies the person name corresponding to the first term by using the personnel DB (first DB) of the management unit 32, and identifies the target person indicated by the first term based on the specified person name. Consists of the identification unit.

Under such a premise, there may be a case where the target person cannot be narrowed down to one person by the person name corresponding to the first term. In other words, there may be a case where there are two or more candidates for the target person narrowed down by the person name corresponding to the first term. In such a case, the extraction unit 33A and the identification unit 33B may perform the following operations.

In option 1, the extraction unit 33A may extract the second term in which the distance from the first term appears within a predetermined distance in the text document. The second term may be extracted by a known morphological analysis or the like. The second term may be referred to as a peripheral term. The distance and the predetermined distance may be defined by the number of characters, the number of words, the number of clauses, or the number of sentences.

The identification unit 33B uses the personnel DB (second DB) of the management unit 32 to identify the affiliation name corresponding to the second term, and based on the specified affiliation name, the target person specified based on the person name. Identify the affiliation of the candidate. The identification unit 33B identifies the target person based on the specified affiliation from the candidates of the target person identified based on the person name.

In option 2, the identification unit 33B uses the personnel DB (third DB) of the management unit 32 to identify the correlation between the candidate of the target person specified based on the person name and the writer. The identification unit 33B identifies the target person based on the identified correlation from the candidates of the target person identified based on the person name. For example, the identification unit 33B may identify the person having the closest correlation distance with the writer as the target person.

In option 3, the extraction unit 33A may extract the second term in which the distance from the first term appears within a predetermined distance in the text document. The second term may be extracted by a known morphological analysis or the like. The second term may be referred to as a peripheral term. The distance and the predetermined distance may be defined by the number of characters, the number of words, the number of clauses, or the number of sentences.

The identification unit 33B uses the personnel DB (4th DB) of the management unit 32 to identify the characteristic words corresponding to the second term by using the database that stores the characteristic words related to the person. The identification unit 33B identifies the target person based on the specified characteristic word from the candidates of the target person specified based on the person name.

Here, two or more options selected from the above-mentioned option 1 (affiliation name), option 2 (correlation), and option 3 (characteristic word) may be combined. For example, the score of each candidate of the target person may be calculated based on option 1 (affiliation designation). The score of each candidate of the target person may be calculated based on option 2 (correlation). The score of each candidate of the target person may be calculated based on option 3 (characteristic word). The score may be considered as an index showing the possibility that the candidate of the target person is the target person. The identification unit 33B may identify the candidate of the target person having the highest total score based on two or more options selected from the options 1 to 3 as the target person.

In such cases, valuations may be applied to the scores based on Options 1 to Option 3. The weighted value is a value for determining the priority of options 1 to 3.

For example, the identification unit 33B may identify the target person by preferentially using the specified affiliation (that is, option 1) when the person designation is the first designation. By "priority" may mean that option 1 takes precedence over option 2 and option 3. "Priority" may mean that the valuation applied to option 1 is greater than the valuation applied to options 2 and 3. In such a case, since the first term applied to the text document is the first name such as name or first name, it is highly probable that the text document is official, so even if the second option is preferentially used. good.

Alternatively, the identification unit 33B may identify the target person by preferentially using the correlation (that is, option 2) when the person designation is the second designation. By "priority" may mean that option 2 takes precedence over option 1 and option 3. "Priority" may mean that the valuation applied to option 2 is greater than the valuation applied to options 1 and 3. In such a case, since the first term applied for in the text document is the second name such as a nickname, it is highly probable that the person close to the writer is the target person, so the second option is preferentially used. You may.

However, these examples are only examples. For example, in the identification unit 33B, even if the person name is the first name, if the person name is only the name, the person close to the writer is the target person. Since the probability is high, the second option may be used preferentially.

(Concrete example)
Hereinafter, specific examples of the embodiments will be described. In the following, the method of identifying the target person indicated by "Mr. AA" will be explained by taking as an example the text document "Please consult with Mr. AA of the Legal Office regarding patent matters". The author of the text document is "PP QQ".

First, as shown in FIG. 4, the extraction unit 33A extracts terms such as "patent", "legal office", "AA", and "consultation" by morphological analysis of a text document. "AA" is an example of the first term. "Patent", "legal office" and "consultation" are examples of the second term (peripheral language).

Second, as shown in FIG. 5, the identification unit 33B identifies and identifies the person name "AA" corresponding to the first term "AA" by using the personnel DB (first DB) of the management unit 32. Attempts to identify the target person indicated by the first term based on the person name given. Here, as illustrated in FIG. 3, three people, "AA XX", "AA YY", and "AA ZZ", are identified as candidates for the target person, and the target person cannot be identified only by the person name. Is illustrated.

When the above-mentioned option 1 is used, the identification unit 33B uses the personnel DB (second DB) of the management unit 32 as shown in FIG. The "legal office" is specified, and the affiliation of the candidate of the target person specified based on the person's name is specified based on the specified affiliation name "legal office". The identification unit 33B identifies the target person based on the specified affiliation from the candidates of the target person identified based on the person name. Here, "AA YY" belonging to the Legal Office may be identified as the target person.

Alternatively, as a score based on option 1, a score "1" of a candidate for a target person having a specified affiliation name (here, "AA YY") is calculated, and a target having no specified affiliation name. A score "0" of a person candidate (here, "AA XX", "AA-ZZ") may be calculated.

When the above-mentioned option 2 is used, the identification unit 33B uses the personnel DB (third DB) of the management unit 32 as a candidate for the target person specified based on the person name, as shown in FIG. Identify the correlation with the writer. The identification unit 33B identifies the target person based on the identified correlation from the candidates of the target person identified based on the person name. Here, "AA YY", which has the closest correlation distance to the writer "PP QQ", may be identified as the target person.

Alternatively, a score based on the correlation distance may be calculated as a score based on option 2. The score based on the correlation distance may be expressed as 1 / (1 + correlation distance). The correlation distance may be defined by the number of paths between the writer and the candidate target person. For example, the score for "AA YY" is 0.5 (= 1 / (1 + 1)), the score for "AA ZZ" is 0.33 (= 1 / (1 + 2)), and "AA X X". The score of may be 0.25 (= 1 / (1 + 3)).

When the above-mentioned option 3 is used, the identification unit 33B uses the personnel DB (4th DB) of the management unit 32 as shown in FIG. 8, and the feature word “patent” corresponding to the second term “patent” is used. To identify. The identification unit 33B identifies the target person based on the specified characteristic word "patent" from the candidates of the target person specified based on the person name. Here, "AA YY" having "patent" as a characteristic word may be identified as the target person.

Alternatively, as a score based on option 3, the score "1" of the candidate of the target person having the specified characteristic word (here, "AA YY") is calculated, and the target having no specified characteristic word is calculated. A score "0" of a person candidate (here, "AA XX", "AA-ZZ") may be calculated.

Here, when options 1 to 3 are combined, as shown in FIG. 9, the total score based on options 1 to 3 is calculated, and the candidate for the target person having the largest total score (here, the candidate). , "AA YY") may be identified as the target person.

Further, as described above, valuations may be applied to the scores based on Options 1 to Option 3. The weighted value is a value for determining the priority of options 1 to 3. For example, when the person designation is the first designation, the weighting value applied to option 1 may be larger than the weighting value applied to options 2 and 3. Alternatively, when the person designation is the second designation, the weighting value applied to option 2 may be larger than the weighting value applied to options 1 and 3.

(Entity identification method)
Hereinafter, the entity identification method according to the embodiment will be described. Here, a case where there are two or more candidates for the target person narrowed down by the person designation corresponding to the first term, and the target person is identified using options 1 to 3 will be described.

As shown in FIG. 10, in step S10, the entity identification device 30 extracts the first term and the second term (peripheral words) by morphological analysis of a text document or the like.

In step S12, the entity identification device 30 identifies the person designation corresponding to the first term by using the personnel DB (first DB) of the management unit 32.

In step S14, the entity identification device 30 identifies the affiliation name corresponding to the second term by using the personnel DB (second DB) of the management unit 32, and based on the identified affiliation name, based on the person name. Identify the affiliation of the identified candidate for the target person (option 1).

In step S16, the entity identification device 30 uses the personnel DB (third DB) of the management unit 32 to identify the correlation between the candidate of the target person specified based on the person name and the writer (option 2). ..

In step S18, the entity identification device 30 identifies the feature word corresponding to the second term by using the personnel DB (fourth DB) of the management unit 32 (option 3).

In step S20, the entity identification device 30 identifies the target person based on the scores based on options 1 to 3 from the candidate target persons identified based on the person designation. Valuations may be applied to scores based on Options 1-3.

(Action and effect)
In an embodiment, the entity identification device 30 extracts a first term relating to a person appearing in a text document, identifies a person designation corresponding to the first term, and the first term indicates based on the identified person designation. Identify the target person. According to such a configuration, the target person (entity) that appears in the text document can be appropriately identified.

In the embodiment, the entity identification device 30 identifies the target person based on one or more options selected from the options 1 to 3 from the target person candidates identified based on the person designation. May be good. According to such a configuration, even if there are two or more target person candidates narrowed down by the person name corresponding to the first term, it is possible to appropriately identify the target person (entity) that appears in the text document. can.

In embodiments, weighted values may be applied to scores based on Options 1 to Option 3. With such a configuration, the priority of options 1 to 3 can be appropriately set. Therefore, the target person (entity) that appears in the text document can be appropriately identified.

[Other embodiments]
Although the invention has been described by embodiments described above, the statements and drawings that form part of this disclosure should not be understood to limit the invention. This disclosure will reveal to those skilled in the art various alternative embodiments, examples and operational techniques.

Although not specifically mentioned in the above disclosure, options 1 to 3 may not be used when the target person can be narrowed down to one person by the person name corresponding to the first term.

In the above disclosure, the case where the target person appearing in the text document is an employee is mainly illustrated. However, the embodiments are not limited to this. The target person may be a person whose first name, second name, affiliation name, correlation, characteristic word, etc. are managed by the management unit 32. For example, the target person may include a person who belongs to a corporate trader.

Although not specifically mentioned in the above disclosure, the person designation may include a personal pronoun. Personal pronouns may be treated in the same way as the second designation described above.

Although not specifically mentioned in the above disclosure, "equivalent" is used in "personal name corresponding to the first term", "affiliation name corresponding to the second term", "characteristic name corresponding to the second term", etc. It may be a term meaning an exact match between the two, or it may be a term meaning a partial match between the two. Alternatively, "equivalent" may mean that the similarity between the two is equal to or greater than the threshold value when the similarity between the two can be scored. For scoring the similarity, a known method such as scoring using the Eugrid distance of each term and scoring using the cosine similarity of each term may be used.

In the above-mentioned disclosure, the case where the 1st DB to the 4th DB are provided in the entity identification device 30 is illustrated. However, the above disclosure is not limited to this. The 1st DB to the 4th DB may be included in the same external server that can communicate with the entity identification device 30, and may be included in a separate external server that can communicate with the entity identification device 30 as a separate body from the entity identification device 30. It may be. At least one of the first DB to the fourth DB may be realized by SaaS (Service as a Software).

Although not specifically mentioned in the above-mentioned disclosure, a program that causes a computer to execute each process performed by the entity identification device 30 may be provided. The program may also be recorded on a computer-readable medium. Computer-readable media can be used to install programs on a computer. Here, the computer-readable medium on which the program is recorded may be a non-transient recording medium. The non-transient recording medium is not particularly limited, but may be, for example, a recording medium such as a CD-ROM or a DVD-ROM.

Alternatively, a chip composed of a memory for storing a program for executing each process performed by the entity identification device 30 and a processor for executing the program stored in the memory may be provided.

10 ... 1st terminal, 20 ... 2nd terminal, 30 ... entity identification device, 31 ... communication unit, 32 ... management unit, 33 ... control unit, 100 ... entity identification system, 200 ... network

Claims (7)

An extractor that extracts the first term related to a person appearing in a text document,
Using the database that stores the names related to the person, the person names corresponding to the first term are identified, and the candidate of the target person identified based on the specified person names is based on the information about the text document. An entity identification device comprising an identification unit for identifying a target person indicated by the first term. The extraction unit extracts the second term in which the distance from the first term appears within a predetermined distance in the text document.
The identification unit is
Using a database that stores names related to the affiliation of a person, the affiliation name corresponding to the second term is specified, and based on the specified affiliation name, the affiliation of the candidate of the target person specified based on the person name. Identify and
The entity identification device according to claim 1, wherein the target person is identified based on the specified affiliation from the candidates of the target person specified based on the person designation. The identification unit according to claim 2, wherein the identification unit identifies the target person by preferentially using the specified affiliation when the person designation is the first designation including at least a part of the name and the name. Entity identification device. The text document is a document that can identify the writer of the text document.
The identification unit is
Using a database that stores the correlations related to the correlation of people, the correlation between the candidate of the target person specified based on the person's name and the writer is identified.
The entity identification device according to any one of claims 1 to 3, wherein the target person is identified based on the specified correlation from the candidates of the target person specified based on the person designation. .. The identification unit claims to identify the target person by preferentially using the specified correlation when the person name is a second name other than the name including at least a part of the name and the name. 4. The entity identification device according to 4. The extraction unit extracts the second term in which the distance from the first term appears within a predetermined distance in the text document.
The identification unit is
Using a database that stores characteristic words related to a person, identify the characteristic words that correspond to the second term, and then identify them.
The entity identification device according to any one of claims 1 to 5, which identifies the target person based on the specified feature word from the candidates of the target person specified based on the person designation. .. The step by which the entity identification device extracts the first term for the person appearing in the text document,
A step in which the entity identification device identifies a person's name corresponding to the first term by using a database for storing a person's name.
The entity identification device comprises a step of identifying a target person indicated by the first term from among candidates for a target person identified based on a specified person designation, based on information about the text document . , Entity identification method.

Images