JP7273888B2 - Decision device, decision method and decision program - Google Patents

Description

The present invention relates to a determination device, a determination method, and a determination program.

Conventionally, there is known a technique called a knowledge base in which concepts such as people and events are treated as entities and relationships between the entities are structured. Also, in order to efficiently create such a knowledge database, there is known a technique of clustering entities and updating relationships between entities based on the clustering results.

Japanese Patent No. 6088091

“On Emerging Entity Detection” Michael Farber, Achim Rettinger, Boulos El Asmar

However, there are cases where it cannot be said that the above-described technology provides a useful knowledge database.

For example, as a method of registering entities in a knowledge database, a method of extracting new entities and relationships between entities from content posted on a network such as the Internet and registering the extracted entities and relationships can be considered. be done. However, when content on a network is used, there is a risk of registering incorrect information as relationships between entities.

The present application has been made in view of the above, and an object thereof is to provide a useful knowledge database.

A determination device according to the present application includes an acquisition unit that acquires, from content, a first target and a second target that have a predetermined relationship as candidates for registration in a predetermined database, and information about the content from which the candidates are obtained. and a determination unit that determines a candidate to be registered among the candidates based on the above.

According to one aspect of an embodiment, a useful knowledge database can be generated.

FIG. 1 is a diagram illustrating an example of determination processing performed by an information providing device according to an embodiment. FIG. 2 is a diagram illustrating an example of search processing performed by the information providing apparatus according to the embodiment; FIG. 3 is a diagram illustrating an example of the flow of processing executed by the information providing device according to the embodiment; FIG. 4 is a diagram illustrating a configuration example of an information providing apparatus according to the embodiment; FIG. 5 is a diagram illustrating an example of information registered in an entity database according to the embodiment; FIG. 6 is a diagram illustrating an example of information registered in a triple database according to the embodiment; FIG. 7 is a diagram illustrating an example of information registered in a reliability table according to the embodiment; FIG. 8 is a flowchart showing an example of the flow of determination processing executed by the information providing device according to the embodiment. FIG. 9 is a flowchart illustrating an example of the flow of search processing executed by the information providing device according to the embodiment. FIG. 10 is a diagram illustrating an example of a hardware configuration;

Embodiments for carrying out the determining device, determining method, and determining program according to the present application (hereinafter referred to as "embodiments") will be described in detail below with reference to the drawings. Note that the determination device, determination method, and determination program according to the present application are not limited to this embodiment. Further, each embodiment can be appropriately combined within a range that does not contradict the processing contents. Also, in each of the following embodiments, the same parts are denoted by the same reference numerals, and overlapping descriptions are omitted.

[1. About the information providing device]
First, with reference to FIG. 1, an example of processing executed by an information providing device 10, which is an example of a determining device, will be described. FIG. 1 is a diagram illustrating an example of determination processing performed by an information providing device according to an embodiment. With reference to FIG. 1, an example of determination processing for determining new triples to be registered in a knowledge database in which knowledge systemized and organized information is registered will be described as processing executed by the information providing apparatus 10 .

[1-1. Overview of Information Providing Device]
The information providing device 10 is an information processing device capable of communicating with the data server 100 and the terminal device 200 via a predetermined network N such as the Internet (see FIG. 4, for example). etc. Note that the information providing device 10 may be capable of communicating with any number of data servers 100 and terminal devices 200 via the network N. FIG.

The data server 100 is an information processing device that manages various types of information, and is realized by, for example, a server device or a cloud system. For example, the data server 100 stores structured data such as Wikidata in which data having such a structure is registered, and structured data such as an already constructed knowledge database, in which the relationship between information is defined as a structure in advance. In addition, the data server 100 registers semi-structured data in which information indicating attributes is added to various data using tag languages such as HTML (Hyper Text Markup Language) and XML (Extensible Markup Language). It is

Note that the data server 100 may manage structured data or semi-structured data containing information of arbitrary content. For example, the data server 100 may manage data including search query logs and the like in arbitrary searches such as web searches, route searches, searches for transaction targets in electronic shopping malls, map searches, content searches, and the like. Further, for example, the data server 100 may manage news, blogs, SNSs (Social Networking Services), etc. posted on websites. The data server 100 may also manage explanatory content that explains various objects such as people, objects, buildings, content, and events. Further, for example, the data server 100 may manage explanatory content, which is content explaining various objects that can be objects of entities in a knowledge database such as Wikipedia (registered trademark) or a dictionary service via the Internet. .

The terminal device 200 is a terminal device used by a user, and is realized by, for example, a PC (Personal Computer), a server device, various smart devices, and the like. For example, the terminal device 200 has a function of communicating with the information providing device 10 and outputting information as a communication result upon receiving various types of information input from the user.

Based on such a configuration, for example, the information providing device 10 provides various information registered in the knowledge database based on a request from the terminal device 200 . For example, the terminal device 200 transmits a search query for searching the knowledge database to the information providing device 10 . In such a case, the information providing device 10 searches the knowledge database for information corresponding to the search query, and provides the search result to the user terminal.

[1-2. Overview of Knowledge Database]
Here, the knowledge database searched by the information providing apparatus 10 will be described. In the knowledge database, various types of knowledge are registered in a systematized and organized state. For example, the knowledge database registers triples, which are sets of two registered entities and information indicating the relationship between objects indicated by the entities (hereinafter referred to as "relational information").

Here, an entity is information corresponding to various things in the world, such as various things that can be subjects such as people, objects, and buildings in the real world, attributes such as occupation and nationality, and various states and events. In the following description, various things corresponding to entities may be referred to as "objects".

Relationship information is information indicating the relationship between two entities. More specifically, the relationship information is information indicating the relationship between the first object indicated by the first entity and the second object indicated by the second entity. For example, when the first target is a person, the second target is a book, and the first target is the author of the second target, the information providing apparatus 10 uses information such as "author" and "author" as the relational information. registration will be accepted.

An entity in the knowledge database of the information providing device 10 may correspond to any thing or event. For example, the information providing apparatus 10 may manage triples such that the first and second targets are persons, and the first target is the parent or spouse of the second target. Further, the information providing apparatus 10 may manage triples such that the first and second objects are foods, and the first object is ingredients for the second object. That is, the information providing apparatus 10 manages triples indicating arbitrary relationships between arbitrary objects, thereby managing information indicating relationships and events between objects, that is, a knowledge base.

[1-3. About decision processing]
Now consider the process of registering a new triple in the knowledge database. Registering a new triple in such a knowledge database requires not only a character string indicating each target, but also relationship information indicating what kind of relationship the target has with other targets. In order to efficiently add such triples to the knowledge database, a method of extracting triples from content posted on the Internet, such as news articles, is conceivable. However, since such content posted on the Internet includes incorrect information, it may not always be possible to extract correct triples.

Therefore, the information providing device 10 executes the following decision processing. First, the information providing apparatus 10 acquires triple candidates including a first target, a second target, and relationship information indicating the relationship between the targets from content on the network. Subsequently, the information providing apparatus 10 associates the acquired candidate with the acquisition source information corresponding to the content from which the candidate is acquired. Then, the information providing apparatus 10 determines triples to be registered from among the candidates based on the acquisition source information associated with each candidate.

For example, triples extracted from domain content whose accuracy is verified by multiple users, such as Wikipedia, are more accurate than triples extracted from content posted on domains such as personal blogs and SNS. estimated to be high. Therefore, the information providing apparatus 10 links acquisition source information, which is the acquisition source of each triple candidate, to each triple candidate. Then, when the acquisition source information associated with a certain triple candidate indicates a reliable acquisition source, the information providing apparatus 10 registers the triple candidate.

Here, even if triples are extracted from content posted on personal blogs, SNS, etc., when many contents contain triples with the same content, that is, triples showing the same relationship between the same objects , such triples are presumed to indicate the correct information. Therefore, if the information providing device 10 is a triple candidate with the same content and the acquisition source information associated with each triple satisfies a predetermined condition, the information providing device 10 registers the triple with the same content as each candidate triple. . That is, the information providing apparatus 10 determines whether or not the reliability of the content indicated by each triple candidate exceeds a predetermined threshold based on the acquisition source information associated with each triple candidate, If the threshold is exceeded, the triple with the content indicated by each triple candidate is to be registered.

For example, the information providing apparatus 10 associates each triple candidate with a reliability, which is a numerical value indicating the reliability of content from which each triple candidate is obtained. To give a more specific example, the information providing apparatus 10 associates the reliability according to the domain from which each triple candidate is acquired. That is, the information providing apparatus 10 associates each triple candidate with the reliability corresponding to the domain of the acquisition source as acquisition source information, which is information corresponding to the acquisition source of each triple.

In addition, the information providing apparatus 10 identifies a plurality of triple candidates indicating the same relationship between the same objects from each triple candidate. For example, the information providing apparatus 10 uses various techniques such as normalization to generate a first entity that indicates the same first target, a second entity that indicates the same second target, and relationship information that indicates the same relationship. Identify candidate triples containing .

Then, the information providing apparatus 10 determines whether or not the reliability associated with the identified triple candidate satisfies a predetermined condition, and if the reliability satisfies the predetermined condition, the information providing apparatus 10 is the same as the identified triple candidate. The triples that show the same relationship between the subjects are registered candidates. For example, when the total reliability associated with the identified triple candidate exceeds a predetermined threshold, the information providing apparatus 10 sets the triple having the same content as the identified triple candidate as a registration candidate. Even if only one triple candidate is identified, the information providing apparatus 10 may determine the triple candidate to be registered if the reliability exceeds a predetermined threshold.

In this way, the information providing apparatus 10 associates the acquisition source of each triple candidate and the corresponding acquisition source information with each triple candidate. Then, when the acquisition source information associated with the triple candidate indicating the relationship between certain objects satisfies a predetermined condition, the information providing apparatus 10 provides the triple candidate or the same content as the triple candidate. to be registered in the knowledge database. As a result of such processing, the information providing apparatus 10 can register triples with highly reliable contents in the knowledge database. As a result of improving the accuracy of the knowledge database, useful information can be provided to users. can be done.

In the above example, the information providing apparatus 10 associates the triple candidate with the reliability corresponding to the acquisition source domain, but the embodiment is not limited to this. For example, the information providing apparatus 10 may link the acquisition source domain with the triple candidate as the acquisition source information. In such a case, the information providing apparatus 10 may register the candidate triple as long as the domain associated with the candidate triple is set in advance as a highly reliable domain. In addition, even if each domain linked to a plurality of triple candidates indicating the same content is set in advance as a low-reliability domain, the information providing apparatus 10 does not generate triple candidates indicating the same content. exceeds a predetermined threshold value, such triple candidates or triple candidates with the same content may be registered as targets for reliability, assuming that reliability is guaranteed. That is, if the information providing apparatus 10 determines triples to be registered among the triple candidates according to the reliability of the acquisition source of the triple candidates, the information providing apparatus 10 uses arbitrary information as the acquisition source information for the triple candidates. You can tie it.

[1-4. About decision processing]
An example of the flow of processing executed by the information providing apparatus 10 will be described below with reference to FIG. In the following description, a candidate triple obtained from content may be simply referred to as a "triple", and a triple to be registered may be referred to as a "registered triple". Further, in the following description, a plurality of triples including a first entity indicating the same first object, a second entity indicating the same second object, and relationship information indicating the same relationship are referred to as "same triples". It may be described.

First, the information providing device 10 collects various contents from the data server 100 (step S1). For example, the information providing device 10 collects an already constructed knowledge database and various posted contents.

Subsequently, the information providing device 10 acquires triples from each content (step S2). For example, the information providing apparatus 10 acquires triples registered in the knowledge database when the acquisition source content is a knowledge database that has already been constructed. In addition, when the content to be acquired is text-based content such as content posted on an SNS or Wikipedia content, the information providing apparatus 10 extracts triples using morphological analysis, intention analysis, or the like. For example, the information providing apparatus 10 extracts a subject, a predicate, and an object contained in the text, and generates a triple having the subject as the first entity, the object as the second entity, and the predicate as relationship information. It should be noted that various well-known techniques can be adopted for the technique of generating triples from such text, and detailed description thereof will be omitted.

For example, in the example shown in FIG. 1, the information providing apparatus 10 generates a triple # containing entity #1-1A, entity #1-1B, and relationship information #1-1 from content #1-1 belonging to domain #1. Get 1-1. The information providing apparatus 10 also acquires triple #2-1 including entity #2-1A, entity #2-1B, and relationship information #2-1 from content #2-1 belonging to domain #2. The information providing apparatus 10 acquires triple #3-1 including entity #3-1A, entity #3-1B, and relationship information #3-1 from content #3-1 belonging to domain #3. Content #1-1 included in domain #1 is structured data, and content #2-1 included in domain #2 and content #3-1 included in domain #3 are unstructured data. shall be

Subsequently, the information providing device 10 associates the acquisition source information with each triple (step S3). For example, the information providing apparatus 10 associates the triple #1-1 with the domain #1 as the acquisition source information, associates the triple #2-1 with the domain #2 as the acquisition source information, and associates the triple #3-1 with the domain #2 as the acquisition source information. 1 is associated with domain #3 as acquisition source information. In step S3, the information providing apparatus 10 may link the reliability corresponding to each of the domains #1 to #3 as the acquisition source information.

Subsequently, the information providing apparatus 10 identifies triples showing the same relationship from the obtained triples, and determines the triples to be registered based on the total reliability of the acquisition sources indicated by the acquisition source information given to the identified triples. A registered triple is determined (step S4). For example, the information providing apparatus 10 selects, from among the acquired triples, a character string indicating the same object as entity #1-1A (for example, a synonym for entity #1-1A) as the first entity, and entity #1-1B as the first entity. A character string indicating the same object as #1-1 is set as the second entity, and another triple whose relational information is a character string indicating the same relationship as the relational information #1-1, that is, the same triple is retrieved. Similarly, the information providing device 10 searches for triples identical to triple #2-1 and triples identical to triple #3-1.

For example, the information providing device 10 makes sure that the target indicated by entity #2-1A matches the target indicated by entity #3-1A, and the target indicated by entity #2-1B matches the target indicated by entity #3-1B. If the relationship indicated by the relationship information #2-1 and the relationship indicated by the relationship information #3-1 match, even if the character strings of the entities and relationship information are different, the triple #2-1 and the triple # 3-1 is determined to be the same triple.

A more specific example will be described. For example, entity #2-1A included in triple #2-1 is the Japanese notation "Shakespeare", entity #2-2 is the Japanese notation "King Lear", and relational information #1-1 is It is assumed that the author is written in Japanese. In addition, the entity #3-1A included in the triple #3-1 is the Japanese notation "William Shakespeare", the entity #2-2 is the Japanese notation "King Lear", and the relationship information #1- Assume that 1 is the English notation "author".

In such a case, the entity “Shakespeare” included in triple #2-1 and the entity “William Shakespeare” included in triple #3-1 have different character strings, but are the same person (so-called British in the 16th century). recognized as indicating Shakespeare, who is a person. Also, the relationship information “author” included in triple #2-1 and the relationship information “author” included in triple #3-1 have different character strings, but have the same relationship (that is, The relationship that the person authored the work indicated by the second entity). Therefore, the information providing apparatus 10 assumes that triple #2-1 and triple #3-1 are triples showing the same relationship, that is, identical triples.

In this way, the information providing apparatus 10 allows triple candidates obtained from different contents to indicate the same relationship between the first entity indicating the same first target and the second entity indicating the same second target. and relationship information, and determined to include a first entity indicating the same first object, a second entity indicating the same second object, and relationship information indicating the same relationship. Whether or not a candidate is to be registered is determined based on the acquisition source information associated with the plurality of triple candidates.

It should be noted that such a process can be realized by referring to, for example, a list that is created in advance and is a list of entity candidates representing the same object. Further, for example, the information providing apparatus 10 can realize such a search for identical triples by referring to an associative memory, a synonym dictionary, or a knowledge database in which synonyms are associated with each other. In addition to such processing, the information providing apparatus 10 may search for identical triples by using various known normalization techniques, for example.

Then, the information providing apparatus 10 determines registered triples based on the acquisition source information of each identical triple. For example, if the triple #1-1 does not have the same triple, the information providing device 10 identifies the trader information "domain #1" associated with the triple #1-1. Here, since the content registered in domain #1 is structured data, it can be considered that a certain degree of reliability is ensured. Therefore, the information providing apparatus 10 gives domain #1 a reliability level of "80", which is higher than a predetermined threshold value (for example, "50").

Also, for example, the information providing device 10 identifies the trading partner information “domain #2” and “domain #3” associated with triple #2-1 and triple #3-1, which are identical triples. Here, since the contents registered in domain #2 and domain #3 are semi-structured data, the reliability of the contents is considered to be lower than that of domain #1. Therefore, the information providing apparatus 10 assigns a reliability level of "50" to the domain #2, and assigns a reliability level of "30" to the domain #3. It should be noted that it is possible to arbitrarily set in advance what kind of reliability is assigned to content of which domain. Giving a higher degree of trust than the content of the data, and assigning a higher degree of trust to the content of the domain whose content is scrutinized by multiple users than to the content of the domain that can be posted by the actions of a single user. may be attached.

Then, the information providing device 10 determines whether or not the total reliability of each identical triple exceeds a predetermined value (for example, "80"). Triples having the same content as these identical triples are registered triples. For example, since the reliability of triple #1-1 is "80", the information providing apparatus 10 sets entity #1-1A as entity #1A, entity #1-1B as entity #1B, and relation information #1. A triple having -1 as relational information #1 is set as registered triple #1. That is, the information providing apparatus 10 sets the triple #1-1 as the registered triple #1.

On the other hand, the information providing apparatus 10 determines that the sum of the reliability of triple #2-1 and the reliability of triple #3-1 is "80". A triple having the same content as that of 3-1 is designated as registered triple #2. For example, the information providing device 10 may be an entity #2A indicating the same target as the entity #2-1A or the entity #3-1A, an entity #2B indicating the same target as the entity #2-1B or the entity #3-1B, And a triple containing relational information #2 indicating the same relationship as relational information #2-1 or relational information #3-1 is defined as registered triple #2. That is, the information providing apparatus 10 generates triples obtained by normalizing the contents of the same triples according to the rules in the knowledge database that is finally generated, and uses the generated triples as registered triples.

For example, when triple #2-1 or triple #3-1 includes the above-described character string, the information providing device 10 includes "William Shakespeare" written in Japanese as entity #2A, and "Japanese text" as entity #2B. A triple containing the word notation "King Lear" and the Japanese notation "author" as relational information #2 is registered triple #2.

Then, the information providing apparatus 10 adds acquisition source information corresponding to the acquisition source to each of the registered triples #1 and #2, and registers them in the knowledge database. For example, the information providing apparatus 10 associates the registered triple #1 with the acquisition source information “domain #1” of the triple #1-1 from which the registered triple #1 is based, and registers it in the knowledge database. Further, the information providing apparatus 10 links the acquisition source information “domain #2” of the original triple #2-1 and the acquisition source information “domain #3” of the triple #3-1 to the registered triple #2. and register it in the knowledge database.

The information providing device 10 also receives a search query from the terminal device 200 (step S5). In such a case, the information providing device 10 searches the knowledge database for information corresponding to the search query, and provides the search result to the terminal device 200 (step S6). At this time, the information providing apparatus 10 uses the acquisition source information associated with each registered triple to narrow down the triples, entities, and related information to be searched.

[1-5. About search processing]
Next, a search process executed by the information providing device 10 will be described with reference to FIG. FIG. 2 is a diagram illustrating an example of search processing performed by the information providing apparatus according to the embodiment; This search processing corresponds to steps S5 and S6 shown in FIG. In the example shown in FIG. 2, as shown in step S4 of FIG. 1, it is assumed that the acquisition source information indicating the acquisition source is linked to the triples in the knowledge database.

For example, the information providing device 10 receives a search query from the terminal device 201 used by the first user (step S1). More specifically, the information providing apparatus 10 receives a search query and designation of a domain to be searched. In such a case, the information providing apparatus 10 searches for information corresponding to the search query using triples associated with the specified domain (step S2).

For example, in the example shown in FIG. 2, domain #1 is associated with registered triple #1 including entity #1 and entity #2, and registered triple #2 including entity #1 and entity #3 is associated with domain #1. are associated with domains #2 and #3, and domain #1 and #2 are associated with registered triple #3 including entity #1 and entity #4. Here, when the information providing apparatus 10 receives "entity #1" and search target "domain #1" as a search query from the first user, among registered triples #1 to #3, "domain From the registered triples #1 and #3 associated with "Entity #1", other entities having a predetermined relationship with "Entity #1" (eg, entity #2, entity #4, etc.) are retrieved. The information providing device 10 then transmits the search result to the terminal device 201 .

On the other hand, for example, when the information providing apparatus 10 receives the search query "entity #1" and the search target "domain #2" from the terminal device 202 used by the second user, the registered triples #1 to # 3, from the registered triples #2 and #3 associated with "domain #2", other entities having a predetermined relationship with "entity #1" (for example, entity #3, entity #4, etc.) search for. The information providing device 10 then transmits the search result to the terminal device 202 .

That is, the information providing apparatus 10 multiplexes the knowledge database for each acquisition source by associating the acquisition source information with the triple. More specifically, the information providing apparatus 10 generates a knowledge database multiplexed for each domain, ie, a multi-domain knowledge database, by associating the acquisition source domain with the triple. Then, the information providing apparatus 10 searches for information corresponding to the search query received from the user, using triples in which the specified acquisition source and the corresponding acquisition source information are associated.

In the example shown in FIG. 2, the information providing apparatus 10 associates the domain from which the triples are acquired as the supplier information, but the embodiment is not limited to this. For example, the information providing device 10 may associate reliability with each triple. More specifically, the information providing apparatus 10 assigns the trust level "80" corresponding to the domain #1 to the registered triple #1, and assigns the trust level corresponding to the domain #2 to the registered triple #2. The trust level "80", which is the sum of the trust level "50" and the trust level "30" corresponding to domain #3, is given to registered triple #3, and the trust level "80" corresponding to domain #1 and domain #1 are added to registered triple #3. A trust of "130" is given, which is the sum of domain #2 and the corresponding trust of "50".

Then, for example, when the reliability specified by the user is "80", the information providing apparatus 10 searches for the registered triples #1 and #3 with the reliability of "80" or more, and If the specified reliability is "100", registered triple #3 with a reliability of "100" or higher may be searched. Further, for example, the information providing apparatus 10 associates and registers each triple with the domain that is the acquisition source, and when receiving the specification of the degree of reliability from the user, among the triples, the associated domain is registered. It is also possible to search for triples for which the sum of the specified domain and the corresponding confidence exceeds the specified confidence.

[1-6. Reliability summation method]
In the example described above, the information providing apparatus 10 determines the registered triples depending on whether the sum of the degrees of reliability corresponding to the sources from which the triples are obtained exceeds a predetermined threshold. Here, the information providing apparatus 10 may search for registered triples depending on whether or not the reliability value obtained by considering various weights exceeds a predetermined threshold.

For example, when the weighted sum of the degrees of reliability associated with identical triples exceeds a predetermined threshold, the information providing apparatus 10 may set triples having the same relationship with these identical triples as registered triples. Note that such weights may be adjustable according to arbitrary information, for example. For example, the information providing apparatus 10 may set a higher weight for content that is the source of triple extraction, the older it is registered, the higher the evaluation from other users, or the higher the number of views. In addition, the information providing apparatus 10 may set a smaller weight for content that is the source of triple extraction, the more recent the registration time, the lower the evaluation from other users, or the lower the number of views. . In addition, the information providing apparatus 10 assigns a higher weight to the character string of each entity included in the triple and the character string of the relationship information, and the character string registered in the reference dictionary information. May be set.

Also, for example, it is conceivable that the trust level for each domain is determined not only absolutely but also relatively. Therefore, the information providing apparatus 10 may calculate the weighted sum of the degrees of reliability associated with the same triple using a weight according to the relationship between the contents from which the triples are obtained. For example, when the domain #5 is relatively more reliable than the domain #4, the information providing apparatus 10 treats the triples obtained from the domain #5 more than the triples obtained from the domain #4. A high value of confidence or weight may be given and the total confidence may be calculated. Further, for example, if the domain #6 is relatively more reliable than the domain #5, the information providing apparatus 10 may determine the triples obtained from the domain #6 rather than the triples obtained from the domain #5. A higher value of confidence or weight may be assigned to each of the values, and the total confidence may be calculated. In other words, the information providing apparatus 10 may change the weight used when calculating the total reliability of the same triples according to the relative reliability between the domains from which the same triples are obtained.

Further, the information providing apparatus 10 associates the triples obtained from the structured data with reliability values higher than a predetermined threshold, and the triples obtained from the semi-structured data with values lower than the predetermined threshold. may be associated with the reliability of That is, the information providing apparatus 10 may associate triples obtained from structured data with higher reliability than triples obtained from semi-structured data.

Further, the information providing apparatus 10 may associate the triple with a reliability corresponding to the domain of the acquisition source and the relationship indicated by the relationship information included in the acquired triple. For example, if the relationship information included in the triples acquired from the first domain with high reliability of the information on the book indicates a relationship with high relevance to the book, such as "author", the information providing apparatus 10 may set relatively high The reliability of the value is linked to the triple, and if the relationship information contained in the triple obtained from the first domain indicates a relationship with low relevance to the book, such as "location", the reliability of a relatively low value is assigned. May be tied to triples.

[1-7. Regarding an example of specific processing]
In the example described above, the information providing apparatus 10 associates the triples acquired from the content with the acquisition source information, and determines the registered triples based on the acquisition source information linked to the triples. Here, the information providing apparatus 10 may improve the accuracy of the knowledge database by executing various processes other than the processes described above.

For example, as described above, the web may include content that expresses the same relationship between the same objects using different words. Therefore, the information providing apparatus 10 may perform various normalizations to map each entity and the target indicated by the entity. Further, the information providing apparatus 10 acquires triples not only from structured data but also from semi-structured data in order to ensure the diversity of the knowledge database. (for example, extracting triples from the tagged information). Further, the information providing apparatus 10 may further improve the accuracy of the knowledge database by using the manually input information as a reference.

A specific example of the process of generating the knowledge database by the information providing apparatus 10 will be described below with reference to FIG. FIG. 3 is a diagram illustrating an example of the flow of processing executed by the information providing device according to the embodiment; As shown in FIG. 3, the information providing apparatus 10 executes a series of processes indicated by system S to generate a knowledge database from structured data and structured data.

For example, the system S executed by the information providing apparatus 10 includes Importer S1, Information Extractor S2, Attribute Converter S3, Entity Matcher S4, Entity Connector S5, Id Assigner S6, Additional Data Combiner S7, Entity y Merger S8, Object Converter S9, Attribute Completer S10, Validator S11, and Exporter S12 (hereinafter , and may be collectively referred to as “each processing S1 to S12”). The system S may be hardware including a functional configuration for executing the processes S1 to S12, or may be various models such as a neural network.

An example of each of the processes S1 to S12 will be described below. The Importer S1 imports data of various schemas as entities so as to match the schema of the knowledge database to be finally created. For example, the Importer S1 extracts triples from structured data in which relationships between entities have already been set, and generates data with acquisition source information attached to the extracted triples.

For example, the Importer S1 identifies what kind of relationship the relational information contained in the triples of the structured data indicates, and identifies the relational information indicating the identified relation in the finally generated knowledge database. . For example, ImporterS1 identifies triples that include as relationship information indicating a first entity indicating a subject, a second entity indicating an object, and a predicate. Each entity and relationship information may have different character strings even if they indicate the same object or relationship. As a result, the Importer S1 identifies a plurality of triples with different entity and relationship information character strings as triples indicating the same relationship between the same objects in the real world. Also, the Importer S1 associates the acquired triples with the acquisition source information, and assigns a unique temporary ID to each triple.

Information Extractor S2 obtains triples from semi-structured data such as web sentences. For example, Information Extractor S2 uses techniques such as tag analysis and morphological analysis to obtain triples from semi-structured data. The Information Extractor S2 also compares the entities registered in the knowledge database with the entities included in the acquired triples, and links the entities indicating the same target.

The Attribute Converter S3 converts the class of input data into the class of the knowledge database. For example, the Attribute Converter S3 associates a class assigned to each entity in the structured data from which triples are obtained (that is, information indicating the target attribute indicated by the entity) with the class of each entity in the knowledge database. Refers to the mapping information and converts the class of each entity included in the fetched triples to the class of the knowledge database. For example, the Attribute Converter S3 normalizes character strings of entities indicating the same object, and normalizes relationship information indicating the same relationship. For example, the Attribute Converter S3 acquires a triple obtained by converting the relationship information "auter" in English into the relationship information "author" in Japanese.

Entity Matcher S4 matches entities that indicate the same object and generates a group of identical entities. For example, Entity Matcher S4 compares the entities contained in each triple and identifies entities that are presumed to exhibit the same relationship of the same object. Entity Matcher S4 then assigns a group ID to the specified entity. Note that the Entity Matcher S4 may identify identical triples so that there is no contradiction in the class (attribute) of each entity. Further, when the entity matcher S4 associates entities with no contradiction in attributes, the entities satisfying the clique structure in the graph structure connecting the entities may be regarded as the same entity.

The Entity Connector S5 links groups of entities that are different in class but related to each other. That is, the Entity Connector S5 links groups of entities that are different in class but are presumed to represent the same object.

The Id Assigner S6 associates each entity group with an ID (Identifier) assigned to the entity in the final knowledge database. For example, the Id Assigner S6 identifies an entity that indicates the same object as the grouped entity among the registered entities, and assigns the ID assigned to the identified entity to the grouped entity.

The Additional Data Combiner S7 manually reflects the correction of the ID assigned by the Id Assigner S6. Further, the Additional Data Combiner S7 associates the entity acquired by the Information Extractor S2 with the ID given to the registered entity indicating the same object.

Entity MergerS8 merges multiple entities representing the same object, ie, the same entity, into one entity. At this time, the Entity Merger S8 performs summation of the reliability and calculation of the weighted sum according to the acquisition source information associated with the entity.

The Object Converter S9 associates an object entity among the entities included in the triple with the ID of the registered entity indicating the same target.

Attribute Completer S10 adds new information to the entity. For example, the Attribute Completer S10 adds information based on a preset ontology. The Attribute Completer S10 may also find or add missing triples based on the ontology. It should be noted that it can be realized by any method such as discovery and addition of such triples, rule base, machine learning, and the like.

The Validator S11 deletes or corrects triples with low reliability based on the acquisition source information. For example, the Validator S11 registers triples only when the sum of reliability calculated by the Entity Merger S8 exceeds a predetermined threshold. Note that the Validator S11 may exclude triples that match a manually set blacklist from being registered, regardless of the reliability.

The Exporter S12 registers triples whose reliability exceeds a predetermined threshold in a predetermined format in the knowledge database. It should be noted that the Exporter S12 may delete information other than the preset information among the information given to each entity of the triple.

[2. Configuration of Information Providing Device]
An example of the functional configuration of the information providing apparatus 10 described above will be described below. FIG. 4 is a diagram illustrating a configuration example of an information providing apparatus according to the embodiment; As shown in FIG. 4, the information providing device 10 has a communication section 20, a storage section 30, and a control section 40. FIG.

The communication unit 20 is realized by, for example, a NIC (Network Interface Card) or the like. The communication unit 20 is connected to the network N by wire or wirelessly, and transmits and receives information to and from the data server 100 and the terminal device 200 .

The storage unit 30 is implemented by, for example, a semiconductor memory device such as a RAM (Random Access Memory) or a flash memory, or a storage device such as a hard disk or an optical disk. The storage unit 30 also stores an entity database 31, a triple database 32, and a reliability table 33 (hereinafter sometimes collectively referred to as "databases 31 to 33").

An example of information registered in each of the databases 31 to 33 will be described below with reference to FIGS. Information about entities is registered in the entity database 31 . For example, FIG. 5 is a diagram showing an example of information registered in the entity database according to the embodiment. As shown in FIG. 5, the entity database 31 registers information having items such as "entity ID", "entity type", "node ID", "node type", and "data".

Here, "entity ID" is an identifier of an entity. The "entity type" is information indicating the type of entity indicated by the associated "entity ID". For example, whether the entity is an entity indicating "person" or an entity indicating "occupation". information such as "Node ID" is an identifier of a node associated with the entity indicated by the associated "Entity ID". The "node type" is information indicating the type of the node indicated by the associated "node ID", and is information such as whether the node indicates a name, a photograph, or an occupation. "Data" is the data of the node indicated by the associated "node ID".

For example, in the example shown in FIG. 5, entity ID "E11", entity type "person", node ID "I111", node type "name", and data "name #1" are associated and registered. Such information indicates that the entity indicated by the entity ID "E11" (that is, the entity E11) is an entity indicating "person", and the node ID "I111" is registered as a node indicating the "name" of the person indicated by the entity. and its name is "Name #1". In the example shown in FIG. 5, conceptual values such as "name #1" and "photo #1" are described, but in reality, the entity database 31 stores the associated entity and the corresponding person Various information indicating the name, photo, date of birth, etc. of the person is registered.

Information indicating triples is registered in the triple database 32 . For example, FIG. 6 is a diagram showing an example of information registered in the triple database according to the embodiment. In the example shown in FIG. 6, the triple database 32 has items such as "triple ID", "relationship information ID", "type", "first element", "second element", and "acquisition source information". Information is registered.

Here, "triple ID" is an identifier for identifying a triple. A “relationship information ID” is an identifier that identifies relational information included in a triple. "Type" is information indicating the relationship between entities included in the triple. Also, the “first element” and the “second element” are the entity IDs of the entities included in the triple indicated by the associated “triple ID”. Also, the “acquisition source information” is information corresponding to the acquisition source of the associated triples, such as information indicating the domain of the acquisition source and information indicating reliability.

For example, in the example shown in FIG. 6, the triple ID "triple #1", the relationship information ID "C1", the type "occupation", the first element "E11", the second element "E21", and the "domain #1" It is associated and registered. In such information, entity E11, entity E12, and relationship information C1 are associated and registered as a triple indicated by triple ID "triple #1", indicating that entity E21 is the occupation of entity E11. Such information also indicates that this triple is information obtained from "domain #1".

In the reliability table 33, reliability information corresponding to the acquisition source of the triple is registered. For example, FIG. 7 is a diagram showing an example of information registered in the reliability table according to the embodiment. As shown in FIG. 7, in the reliability table 33, "domain" and "reliability" are associated and registered. Here, "domain" is information indicating a domain from which triples are obtained. For example, in the example shown in FIG. 7, the domain "domain #1" and the reliability "80" are associated and registered. Such information indicates that the reliability of the triple obtained from the content of the domain indicated by the domain "domain #1" is "80".

Returning to FIG. 4, the description is continued. The control unit 40 is a controller, and various programs stored in a storage device inside the information providing apparatus 10 are transferred to a RAM or the like by a processor such as a CPU (Central Processing Unit) or an MPU (Micro Processing Unit). is executed as a work area. Also, the control unit 40 is a controller, and may be implemented by an integrated circuit such as an ASIC (Application Specific Integrated Circuit) or an FPGA (Field Programmable Gate Array).

As shown in FIG. 2 , the control unit 40 has an acquisition unit 41 , a linking unit 42 , a determination unit 43 , a determination unit 44 , a registration unit 45 and a search unit 46 . Although the functional configuration shown in FIG. 4 does not necessarily match the system S shown in FIG. 3, the system shown in FIG. Similarly, by registering registered triples in the knowledge database according to the reliability of the triples, the accuracy of the knowledge database can be improved. In other words, the functional configuration shown in FIG. 4 shows the main functions exhibited by the processes S1 to S12 shown in FIG.

The acquisition unit 41 acquires triple candidates including a first target, a second target, and relationship information indicating the relationship between the targets from content on the network. For example, the acquisition unit 41 acquires structured data or semi-structured data from the data server 100, and from the acquired content, the first entity, the second entity, and the relationship indicating the relationship between the objects indicated by each entity. Extract triples that are pairs with information. Note that the acquisition unit 41 may acquire triples from structured data in which relationships between objects are defined in advance, or from semi-structured data in which data attributes are set, as content on the network. .

The associating unit 42 associates the acquisition source information corresponding to the acquisition source content with the triple acquired by the acquisition unit. For example, the linking unit 42 links information indicating the content from which the triple was acquired, information indicating the domain to which the content belongs, or reliability according to the content or domain from which the triple was acquired, to the triple as the acquisition source information. wear. Note that, for example, the linking unit 42 links triples obtained from the structured data with a reliability value higher than a predetermined threshold, and triple candidates obtained from the semi-structured data have a reliability higher than the predetermined threshold. may be associated with lower confidence values.

The determination unit 43 determines that triple candidates obtained from different contents include a first entity indicating the same first object, a second entity indicating the same second object, and relationship information indicating the same relationship. Determine whether or not For example, the determination unit 43 uses various dictionaries, ontologies, and the like to determine whether or not each entity and relationship information included in the triples acquired by the acquisition unit 41 indicate the same object or relationship. That is, the determination unit 43 determines whether each triple is the same triple.

The determination unit 44 determines a triple to be registered from among the candidates based on the acquisition source information associated with each triple. For example, if the reliability associated with the triple satisfies a predetermined condition, the determination unit 44 sets the candidate triple as a candidate for registration. In addition, when the sum of reliability or the sum of weights associated with the same triple exceeds a predetermined threshold value, the determination unit 44 may register the triple showing the same relationship of the same target as the same triple. That is, the determination unit 44 determines that the first target indicated by the first entity, the second target indicated by the second entity, and the relationship indicated by the relationship information are the same among a plurality of triples obtained from different contents. If the reliability associated with each of a plurality of triples satisfies a predetermined condition, triples having the same content as each triple may be registered.

Note that the determination unit 44 may calculate the sum of weights using weights according to the relevance of the content that is the acquisition source. For example, the determination unit 44 may apply a weight corresponding to the relative reliability between the domains of the content that is the acquisition source to the weighted sum. Further, for example, when acquisition source information indicating an acquisition source domain is associated with each triple, the determining unit 44 reads the reliability associated with the acquisition source information from the reliability table 33, and reads out the reliability. Whether or not the triple is to be registered may be determined based on whether or not the reliability obtained satisfies a predetermined condition. For example, the acquisition source indicated by the acquisition source information and the number of identical Whether or not a triple is to be registered may be determined based on whether or not a predetermined condition is satisfied.

The registration unit 45 associates the triples to be registered with the acquisition source information and registers them in the database. For example, the registration unit 45 registers in the entity database 31 the information of each entity included in the triple to be registered. Further, the registration unit 45 associates the entity ID, relationship information, and acquisition source information of each entity included in the triple with each other and registers them in the triple database 32 .

The search unit 46 searches for information corresponding to the search query according to the acquisition source information associated with each triple. For example, the search unit 46 receives designation of a domain together with a search query from the terminal device 200 . In such a case, the search unit 46 identifies, from the triple database 32, a triple associated with the domain specified by the terminal device 200 as the acquisition source information, and searches the data of each entity included in the identified triple. , to find the search query and corresponding information.

Note that, for example, when the reliability is received from the terminal device 200, the search unit 46 refers to the triple database 32 and the reliability table 33, and the reliability indicated by the linked acquisition source information is received from the terminal device 200. Identify triples with higher than confidence. Then, the search unit 46 may search for information corresponding to the search query, with the data of each entity included in the identified triple as the search target. After that, the search unit 46 provides the terminal device 200 with the search result.

[3. Example of flow of processing executed by information providing device]
Next, the flow of determination processing executed by the information providing device 10 will be described with reference to FIG. FIG. 8 is a flowchart showing an example of the flow of determination processing executed by the information providing device according to the embodiment.

First, the information providing apparatus 10 acquires triples from each content (step S101). Subsequently, the information providing apparatus 10 associates the acquisition source information corresponding to the acquisition source of the triple with the triple (step S102). Further, the information providing device 10 identifies triples showing the same relationship (step S103). Then, the information providing apparatus 10 determines whether or not the triples indicating the relationships indicated by these triples are to be registered, based on the acquisition source information associated with the identified triples (step S104). After that, the information providing apparatus 10 registers the triple to be registered in the database in a state of being associated with the acquisition source information (step S105), and ends the process.

Next, with reference to FIG. 9, the flow of search processing executed by the information providing device 10 will be described. FIG. 9 is a flowchart illustrating an example of the flow of search processing executed by the information providing device according to the embodiment.

First, the information providing apparatus 10 determines whether or not a search query has been received from the terminal device 200 (step S201), and if not (step S201: No), waits for processing. On the other hand, when a search query is received (step S201: Yes), the information providing apparatus 10 identifies a triple corresponding to the acquisition source indicated by the search query (step S202). Then, the information providing apparatus 10 traces the identified triples and searches for entities corresponding to the search query (step S203). After that, the information providing device 10 provides the search result (step S204) and terminates the process.

[4. Modification]
An example of the determination process and the search process by the information providing apparatus 10 has been described above. However, embodiments are not so limited. Variations of determination processing and search processing executed by the information providing apparatus 10 will be described below.

[4-1. About the type of entity]
In the above example, as an example of the knowledge database, triples and the like indicating the relationship between a person and a work have been described, but embodiments are not limited to this. In other words, the information providing device 10 may update the knowledge database indicating arbitrary matters.

[4-2. Device configuration〕
The information providing device 10 may not manage the knowledge database by itself. For example, each database 31 to 33 registered in the storage unit 30 may be held in an external storage server. Further, the information providing apparatus 10 may be implemented by a front-end server that implements search processing and a back-end server that implements determination processing. In such a case, the units 41 to 45 shown in FIG. 4 are arranged in the back-end server, and the search unit 46 for performing the function of searching the knowledge database is arranged in the front-end server.

[4-3. others〕
Further, among the processes described in the above embodiments, all or part of the processes described as being automatically performed can be manually performed, and conversely, the processes described as being performed manually can be performed manually. can also be performed automatically by known methods. In addition, information including processing procedures, specific names, various data and parameters shown in the above documents and drawings can be arbitrarily changed unless otherwise specified. For example, the various information shown in each drawing is not limited to the illustrated information.

Also, each component of each device illustrated is functionally conceptual, and does not necessarily need to be physically configured as illustrated. In other words, the specific form of distribution and integration of each device is not limited to the one shown in the figure, and all or part of them can be functionally or physically distributed and integrated in arbitrary units according to various loads and usage conditions. Can be integrated and configured.

Moreover, each of the embodiments described above can be appropriately combined within a range that does not contradict the processing contents.

[4-4. program〕
Also, the information providing apparatus 10 according to the above-described embodiment is implemented by a computer 1000 configured as shown in FIG. 10, for example. FIG. 10 is a diagram illustrating an example of a hardware configuration; A computer 1000 is connected to an output device 1010 and an input device 1020, and an arithmetic device 1030, a primary storage device 1040, a secondary storage device 1050, an output IF (Interface) 1060, an input IF 1070, and a network IF 1080 are connected via a bus 1090. have

The arithmetic device 1030 operates based on programs stored in the primary storage device 1040 and the secondary storage device 1050, programs read from the input device 1020, and the like, and executes various processes. The primary storage device 1040 is a memory device such as a RAM that temporarily stores data used by the arithmetic device 1030 for various calculations. The secondary storage device 1050 is a storage device in which data used for various calculations by the arithmetic device 1030 and various databases are registered. It is realized by

The output IF 1060 is an interface for transmitting information to be output to the output device 1010 that outputs various types of information such as a monitor and a printer. It is realized by a connector conforming to a standard such as HDMI (registered trademark) (High Definition Multimedia Interface). Also, the input IF 1070 is an interface for receiving information from various input devices 1020 such as a mouse, keyboard, scanner, etc., and is realized by, for example, USB.

Note that the input device 1020 includes, for example, optical recording media such as CDs (Compact Discs), DVDs (Digital Versatile Discs), PDs (Phase change rewritable discs), magneto-optical recording media such as MOs (Magneto-Optical discs), and tapes. It may be a device that reads information from a medium, a magnetic recording medium, a semiconductor memory, or the like. Also, the input device 1020 may be an external storage medium such as a USB memory.

Network IF 1080 receives data from other devices via network N and sends the data to arithmetic device 1030, and also transmits data generated by arithmetic device 1030 via network N to other devices.

The arithmetic device 1030 controls the output device 1010 and the input device 1020 via the output IF 1060 and the input IF 1070 . For example, arithmetic device 1030 loads a program from input device 1020 or secondary storage device 1050 onto primary storage device 1040 and executes the loaded program.

For example, when the computer 1000 functions as the information providing device 10 , the arithmetic device 1030 of the computer 1000 implements the functions of the control unit 40 by executing programs loaded on the primary storage device 1040 .

[5. effect〕
As described above, the information providing apparatus 10 acquires triple candidates including the first object, the second object, and the relationship information indicating the relationship between the objects from the content on the network. In addition, the information providing apparatus 10 associates the acquired candidate with the acquisition source information corresponding to the content from which the candidate is acquired. Then, the information providing apparatus 10 determines triples to be registered from among the candidates based on the acquisition source information associated with each candidate. As a result of such processing, the information providing apparatus 10 registers, for example, triples acquired from a highly reliable acquisition source, so that the reliability can be ensured while automatically expanding the knowledge database. As a result, a useful knowledge database can be provided.

Here, the information providing apparatus 10 may associate reliability, which is a numerical value indicating the reliability of the content, as the acquisition source information. Further, the information providing apparatus 10 may link the reliability according to the domain of the content as the acquisition source information. Then, when the reliability associated with the triple candidate satisfies a predetermined condition, the information providing apparatus 10 registers the triple candidate. For example, the information providing apparatus 10 is a plurality of candidates acquired from different contents, and the reliability associated with each of the plurality of candidates that match the first target, the second target, and the relationship information is When a predetermined condition is satisfied, a triple containing the first target, the second target, and the relational information is registered. Note that when the weighted sum of the degrees of reliability associated with a plurality of candidates whose relationship information matches the first target, the second target, and the relationship information exceeds a predetermined threshold, the information providing apparatus 10 determines that the first target, A triple that includes the second target and related information may be registered. As a result of such processing, the information providing apparatus 10 can ensure the reliability of the triples to be registered, and thus can provide a useful knowledge database.

In addition, the information providing apparatus 10 uses a weight according to the relationship between the content from which each candidate is acquired, and associates a plurality of candidates whose first target, second target, and related information match each other. A weighted sum of the reliability may be calculated. As a result, the information providing apparatus 10 calculates the reliability of triples in consideration of the relationship between contents, for example, so that the reliability of triples to be registered can be further improved.

In addition, the information providing apparatus 10 acquires triple candidates from structured data in which relationships between objects are defined in advance, or semi-structured data in which data attributes are set, as content on the network. . Here, the information providing apparatus 10 associates the triple candidate acquired from the structured data with a reliability value higher than a predetermined threshold value, and associates the triple candidate acquired from the semi-structured data with a predetermined Confidence values lower than a threshold value may be associated. Therefore, the information providing apparatus 10 can automatically expand the knowledge database and provide a highly reliable knowledge database.

Further, the information providing apparatus 10 determines that triple candidates obtained from different contents have the same relationship between the first element indicating the same first target, the second element indicating the same second target, and the relationship information indicating the same relationship. and is included. Then, the information providing apparatus 10 selects a plurality of triples determined to include the first element indicating the same first target, the second element indicating the same second target, and the relationship information indicating the same relationship. Based on the acquisition source information associated with the candidate, it is determined whether or not the candidate is to be registered. Therefore, the information providing device 10 can ensure the reliability of triples registered in the knowledge database.

Further, when the acquisition source information satisfies a predetermined condition, the information providing apparatus 10 generates a triple including a first element indicating a first target, a second element indicating a second target, and relationship information indicating a relationship. Subject to registration. Therefore, the information providing device 10 can ensure the reliability of the registered triples.

Further, the information providing apparatus 10 associates the triples to be registered with the acquisition source information and registers them in the database. Therefore, the information providing apparatus 10 can provide, for example, a knowledge database capable of realizing a search in consideration of the source of acquisition and the reliability of the source of acquisition.

As described above, some of the embodiments of the present application have been described in detail based on the drawings. It is possible to carry out the invention in other forms with modifications.

Also, the "section, module, unit" described above can be read as "means" or "circuit". For example, the assigning unit can be read as specifying means or a specifying circuit.

10 information providing device 20 communication unit 30 storage unit 31 entity database 32 triple database 33 reliability table 40 control unit 41 acquisition unit 42 linking unit 43 determination unit 44 determination unit 45 registration unit 46 search unit 100 data server 200, 201, 202 terminal Device

Claims (14)

an acquisition unit that acquires, from content, a first target and a second target having a predetermined relationship as candidates for registration in a predetermined database;
and a determination unit configured to determine candidates to be registered among the candidates based on information about content from which the candidates are acquired. 2. The determining device according to claim 1, wherein the determination unit determines a registration target based on reliability, which is a numerical value indicating reliability of the content, as the information regarding the content that is the acquisition source. 3. The determining device according to claim 2, wherein the determination unit determines a registration target based on reliability according to a domain of the content as the information regarding the content that is the acquisition source. The determination device according to claim 2 or 3, wherein, when the reliability associated with the candidate satisfies a predetermined condition, the determination unit treats the candidate as a registration target. The determination unit includes a plurality of candidates obtained from different contents, the first target, the second target, and relationship information indicating a relationship between the first target and the second target. If the reliability associated with each of a plurality of candidates that match , satisfies a predetermined condition, the triple containing the first target, the second target, and the relevant information is registered. The decision device according to any one of claims 2 to 4, wherein: When the weighted sum of the degrees of reliability associated with a plurality of candidates that match the first target, the second target, and the relationship information exceeds a predetermined threshold, the determination unit determines that the first target and 6. The determining apparatus according to claim 5, wherein a triple including said second target and said related information is registered. The determining unit associates a plurality of candidates with matching first target, second target, and related information using a weight according to the relationship between content from which each candidate is acquired. 7. The decision device according to claim 6, further comprising: calculating a weighted sum of the reliability degrees obtained by The acquisition unit acquires the candidate from structured data in which relationships between objects are defined in advance, or from semi-structured data in which data attributes are set, as the content. Clause 8. A decision device according to any one of clauses 1-7. The determination unit determines that the candidate obtained from the structured data has a reliability value higher than a predetermined threshold, and the candidate obtained from the semi-structured data has a reliability value lower than the predetermined threshold. 9. The determining device according to claim 8, wherein, among the candidates, a candidate whose reliability satisfies a predetermined condition is determined as a candidate to be registered. a determination unit that determines whether or not candidates acquired from different contents include a first element indicating the same first target and a second element indicating the same second target,
The determining unit determines, based on information about content from which a plurality of candidates are acquired that are determined to include a first element indicating the same first target and a second element indicating the same second target, 10. The determining device according to any one of claims 1 to 9, which determines whether or not the candidate is to be registered. The determining unit, when the information about the content as the acquisition source satisfies a predetermined condition, a first element indicating the first target, a second element indicating the second target, and the first target and the target. 11. The determining device according to claim 10, wherein triples including relational information indicating a relationship with the second object are to be registered. 12. The method according to any one of claims 1 to 11, further comprising a registration unit that associates the triples to be registered by the determination unit with information about the content that is the acquisition source and registers the triples in a database. decision device. A decision method executed by a decision device,
an obtaining step of obtaining, from the content, a first target and a second target having a predetermined relationship as candidates for registration in a predetermined database;
and a determining step of determining candidates to be registered among the candidates based on information about content from which the candidates are acquired. an acquisition procedure for acquiring, from content, a first target and a second target having a predetermined relationship as candidates for registration in a predetermined database;
A determination program for causing a computer to execute a determination procedure for determining candidates to be registered among the candidates based on information about content from which the candidates are acquired.

Images