JP2019053469A - Database creating device, database creating method, and program - Google Patents

Abstract

To provide a database creating device, a database creating method, and a program capable of creating a database with a high versatility.SOLUTION: A database creating device includes: a determining unit which compares an attribute of a first object registered in a first database with an attribute of a second object registered in a second database different from the first database, and which determines whether or not the first object and the second object are the same object; and a creating unit that creates a database obtained by unifying the first object with the second object when the determining unit determines that the first object and the second object are the same object.SELECTED DRAWING: Figure 2

Description

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

  2. Description of the Related Art Conventionally, a technique for performing automatic question answering with a knowledge graph based on a deep neural network is known (see, for example, Patent Document 1). A single database such as a knowledge graph is generated exclusively by a crawler referring to a plurality of text data described in HTML (HyperText Markup Language).

Japanese Unexamined Patent Publication No. 2017-76403

  However, in the conventional technique, it is difficult to gather various kinds of knowledge in a single database, and in general, a database is generated limited to a specific field. For this reason, the versatility of the database may be reduced.

  The present invention has been made in view of the above problems, and an object of the present invention is to provide a database generation device, a database generation method, and a program capable of generating a highly versatile database.

  According to an aspect of the present invention, the attribute of the first target thing registered in the first database is compared with the attribute of the second target thing registered in a second database different from the first database. A determination unit that determines whether the target thing and the second target thing are the same target thing, and the first target thing and the second target thing are the same target thing by the determination unit. If it is determined, the database generation device includes a generation unit that generates a database in which the first target thing and the second target thing are integrated.

  According to one embodiment of the present invention, a database generation device, a database generation method, and a program that can generate a highly versatile database can be provided.

It is a figure which shows an example of the information provision system 1 containing the database production | generation apparatus 100 in 1st Embodiment. It is a figure which shows an example of a structure of the database production | generation apparatus 100 in 1st Embodiment. It is a figure which shows an example of 1st database DB1. It is a figure which shows an example of 2nd database DB2. It is a figure for demonstrating the attribute determination of an entity concretely. It is a figure which shows an example of knowledge base KB. It is a figure for demonstrating the integration method of an entity. It is a flowchart which shows an example of the process performed by the control part 110 in 1st Embodiment. It is a figure which shows typically the processing content at the time of associating a similar entity with respect to the attention entity. It is a flowchart which shows an example of the process performed by the control part 110 in 2nd Embodiment. It is a figure which shows an example of the hardware constitutions of the information provision apparatus 10 and database production | generation apparatus 100 of embodiment.

  Hereinafter, a database generation device, a database generation method, and a program to which the present invention is applied will be described with reference to the drawings.

[Overview]
The database generation device is realized by one or more processors. The database generation device compares an attribute of a certain target object registered in the first database among a plurality of different databases and an attribute of a certain target object registered in a second database different from the first database. Then, it is determined whether or not these target things are the same target thing. When the database generation device determines that the two target things to be compared are the same target thing, the knowledge type in which the attribute of the other target thing is added to one of the target things. Database (hereinafter referred to as a knowledge base). Thereby, a highly versatile database can be generated.

  The first database or the second database may be, for example, a database that stores information related to locations such as map information, traffic information, and commercial facility information, or a database that stores information related to finance such as stock prices and currency exchange. Alternatively, it may be an encyclopedia database or other various databases.

  A knowledge base is a database that describes information about things and relationships between things as graphs. Things in the knowledge base are, for example, things with the shape of “human”, “machine”, “building”, or abstract concepts such as “organization”, “beauty”, “study”, “travel” (= things) )including. In the present embodiment, among the things described in the knowledge base, the target thing that the database generation apparatus handles in distinction from other things will be described as “entity” in particular.

  An entity (target object) may represent an entity of a target object (for example, an object existing in the real world) or a concept of a target object (for example, in the real world or a virtual world) in the knowledge base. It may represent a concept defined by For example, when the target object is a concept of “building”, the entity may represent an entity such as “XX tower” or “XX building”. Further, for example, when the object is the concept of “economics”, the entity may represent an abstract concept having no entity such as “microeconomics” or “macroeconomics”. In the present embodiment, in order to simplify the description, the entity will be described as representing an entity having a URI (Uniform Resource Identifier) of a content resource on the Internet.

  Things (including entities) described in the knowledge base are defined by an ontology. An ontology is a definition of classes and properties of things, and is a collection of constraints (rules) that hold between classes and properties.

  A class is a group of things that have the same properties in an ontology. The nature of the thing, that is, the class to which the thing belongs is determined by a property (attribute) described later.

  For example, an object that has a beak, is an ovarian vertebrate, and has a wing on its forelimbs, is classified into a class of “bird”. In addition, in the class of “bird”, things that cannot fly are classified into lower classes such as “penguin” and “ostrich”, for example. In this way, the class system may have a hierarchical structure having an upper and lower relationship. The nature of the upper class is inherited by the lower class. In the above example, the nature of the class “bird”, “beaked, ovarian vertebrate, wings on the forelimbs” is the nature of the lower class of “penguins” and “ostriches”. Will also be included. For example, if the class name for identifying the class is a class “bird”, the class name may be represented by a character string “bird”. Note that the class name does not necessarily represent a meaning, and for example, even a class of “bird” may be assigned a character string indicating simple identification information such as “information 1” or “C1”. . The above-described entities, ie, things, belong to any class included in the class system defined by the ontology.

  Properties are attributes that describe the nature and characteristics of things and the relationships between classes. For example, the property may be an attribute indicating the property of “having as a body component” or the property of “having inhabit”, or “a certain class is an upper class and a certain class is a lower class” It may be an attribute indicating the upper / lower relationship between classes, which is “class”. Similar to the class name described above, the property name for identifying the property may be a character string representing the meaning of the property name itself, or may be a character string representing simple identification information.

<First Embodiment>
[overall structure]
FIG. 1 is a diagram illustrating an example of an information providing system 1 including a database generation device 100 according to the first embodiment. The information providing system 1 in the first embodiment includes, for example, one or more information providing apparatuses 10 and a database generation apparatus 100. These devices are connected via a network NW. Note that the information providing apparatus 10 may be aggregated in the database generation apparatus 100.

  Each apparatus shown in FIG. 1 transmits and receives various information via the network NW. The network NW includes, for example, a radio base station, a Wi-Fi access point, a communication line, a provider, the Internet, and the like. Note that it is not necessary for all combinations of the devices shown in FIG. 1 to be able to communicate with each other, and the network NW may partially include a local network.

  The information providing apparatus 10 may be, for example, an application server apparatus that communicates with a terminal device (not shown) in which an application is activated to exchange various types of information, or provides a website such as a search site. It may be a web server device. The terminal device is, for example, a mobile phone such as a smartphone, a computer device such as a tablet terminal or a personal computer.

  For example, the information providing apparatus 10 may extract information corresponding to the search query acquired from the terminal apparatus from the knowledge base generated by the database generation apparatus 100, and provide the extracted information to the terminal apparatus as a search result.

  The database generation apparatus 100 periodically crawls websites provided by the information providing apparatus 10 and websites provided by other web servers, collects content CT on the website, It is a device that collects the managed database itself. The content CT collected by the database generation device 100 includes HTML that is a source of a web page constituting the website, in addition to moving image data, image data, audio data, and text data reproduced or displayed on the website. A document (text data) may be included.

  Then, the database generation device 100 generates a knowledge base KB based on the collected content and database. Details of the method of generating the knowledge base KB will be described later.

[Database generator configuration]
FIG. 2 is a diagram illustrating an example of the configuration of the database generation device 100 according to the first embodiment. As illustrated, the database generation device 100 includes a communication unit 102, a control unit 110, and a storage unit 130, for example.

  The communication unit 102 includes, for example, a communication interface such as a NIC (Network Interface Card) or a DMA (Direct Memory Access) controller. The communication unit 102 communicates with the information providing apparatus 10 and other web servers via the network NW.

  The control unit 110 includes, for example, an acquisition unit 112, a determination unit 114, and a knowledge base generation unit 116. These components are realized, for example, when a processor such as a CPU (Central Processing Unit) executes a program stored in the storage unit 130. Some or all of the components of the control unit 110 are hardware such as LSI (Large Scale Integration), ASIC (Application Specific Integrated Circuit), FPGA (Field-Programmable Gate Array), or GPU (Graphics Processing Unit). It may be realized by (circuit unit; circuit) or may be realized by cooperation of software and hardware.

  The storage unit 130 is realized by, for example, an HDD (Hard Disc Drive), a flash memory, an EEPROM (Electrically Erasable Programmable Read Only Memory), a ROM (Read Only Memory), a RAM (Random Access Memory), or the like. In addition to various programs such as firmware and application programs, the storage unit 130 stores the first database DB1, the second database DB2, the content CT, and the knowledge base KB. Such information will be described later.

  The acquisition unit 112 uses the communication unit 102 to communicate with the information providing apparatus 10 that provides a website and other web servers, and acquires a database managed by these apparatuses.

  For example, the acquiring unit 112 acquires the first database DB1 from the information providing apparatus 10 that provides a certain website A, and receives the second database from the information providing apparatus 10 that provides the website B different from the website A. Get DB2. Then, the acquisition unit 112 stores these databases in the storage unit 130.

  FIG. 3 is a diagram illustrating an example of the first database DB1. For example, the first database DB1 is a database in which information related to locations provided on the website A is accumulated. The URL (Uniform Resource Locator) of each web page is the title of the web page and the commercial indicated by the location. This is information associated with information such as location information such as business hours, fixed holidays, and latitude and longitude of the facility. These pieces of information are provided as content on the web pages that make up the website A.

  FIG. 4 is a diagram illustrating an example of the second database DB2. For example, the second database DB2 is an encyclopedia database provided on the website B. For example, when a dictionary for a certain location is searched for an encyclopedia, the second database DB2 uses the URL of each web page, the title of the web page, the nearest station to that location, and location information. Such information is associated with each other. These pieces of information are provided as content on the web pages that make up the website B.

  Further, for example, when a web page is compatible with a plurality of languages, a title corresponding to each language may be associated with the URL of each web page as in the illustrated example. Also, for example, when a dictionary for a certain person is searched for an encyclopedia, the second database DB2 corresponds to the URL of each web page with the title of the web page, personal information of the person, etc. May be attached. The personal information includes, for example, various information such as a person's name, date of birth, and birthplace.

  In the above-described example, the first database DB1 is a database that stores information related to locations, and the second database DB2 is described as being an encyclopedia database. It may be a database that stores information on finance. In this case, in the first database DB1 or the second database DB2, the URL of each web page is associated with the title of the web page, the name of the company introducing the stock price, etc. on the web page, the security code, etc. It may be information.

  Further, the acquisition unit 112 communicates with the information providing apparatus 10 and other web servers at a predetermined cycle using the communication unit 102, acquires the content CT from the website provided by these apparatuses, and stores the content CT You may memorize | store in the part 130. FIG.

  For example, the acquisition unit 112 transmits a request to each device that provides a website using the communication unit 102 in order to acquire a copy of an HTML document that is newer than the HTML document that has already been acquired as the content CT. Then, when a copy of the HTML document is transmitted from each device, the acquisition unit 112 collects an HTML document of another linked web page from all hyperlinks included in the HTML document. When the acquisition unit 112 accesses a web page of a new HTML document during the collection process, the acquisition unit 112 newly stores the content CT of the web page in the storage unit 130. If a copy of the HTML document is not transmitted by each device until a certain time has elapsed since the transmission of the request, the acquisition unit 112 determines that there is no web page corresponding to the already acquired HTML document, and this web page Content CT collected from the storage unit 130 is deleted.

  The determination unit 114 refers to the first database DB1 and the second database DB2 acquired by the acquisition unit 112, and is registered in each database in order to determine whether there is data that can be integrated into one. It is determined whether the plurality of data satisfy a predetermined determination condition.

  For example, the determination unit 114 extracts information on a predetermined column (field) from information registered in each database as an entity that is a form of information handled in the knowledge base KB, and extracts information as an entity. The various information associated with is used as the attribute of the extracted entity. For example, when a character string described as the title of a web page is extracted as an entity in the first database DB1, a character string input in columns such as URL, business hours, regular holidays, and location information becomes an attribute. The character string includes one or more characters that can be processed by a computer, such as kanji, hiragana, katakana, alphabet, numbers, and symbols. Further, the attribute indicating the name of the entity may be the information itself extracted as the entity. For example, when a character string described as the title of a web page is extracted as an entity, a character string described as the title of the web page (for example, “OO Garden”) is assigned to the name of the entity.

  The determination unit 114 compares the attribute of the entity extracted from the first database DB1 with the attribute of the entity extracted from the second database DB2, and determines whether the comparison target entities are the same entity. For example, the determination unit 114 determines whether or not the comparison target entities are the same entity by determining whether or not various determination conditions as exemplified below are satisfied.

Condition (1): The URL of a predetermined web page matches the character string described as the attribute of the URL of the web page.
Condition (2): The attribute indicating the Japanese title of the web page matches.
Condition (3): The attribute indicating the English title of the web page matches.
Condition (4): The position information and the attribute indicating the name match.

  FIG. 5 is a diagram for specifically explaining entity attribute determination. In the figure, e1, e2, e3,... Represent entities.

  When the database is acquired by the acquisition unit 112, the determination unit 114 extracts information on predetermined columns of each database as an entity, and refers to the URL attribute of the web page among the attributes of the extracted entity. Then, it is determined whether or not the character string described as the URL attribute matches the URL of a predetermined web page (step S100 in the figure).

  The URL of the predetermined web page is the URL of the web page where the information included in the first database DB1 or the second database DB2 is provided. That is, the URL of the predetermined web page is the URL of one web page that constitutes the above-described website A or website B. “Character strings match” may mean that the character strings to be compared are completely matched, or match up to a certain degree of similarity, such as matching up to 8 of 10 characters. It may be to do.

  For example, when the determination unit 114 compares a URL as an attribute with a URL of a predetermined web page, a common website such as a scheme, a host name, and a domain name among character strings constituting the URLs. If the character strings described in the same content are at least matched, the character string described as the URL attribute may be determined to match the URL of a predetermined web page. That is, the determination unit 114 compares the attributes of the entities to determine whether the website from which the data extracted as the entities is acquired is the same site. The determination unit 114 collects entities determined to have the character string described as the URL attribute matching the URL of the predetermined web page into one cluster CL. At this time, the determination unit 114 may combine entities having no URL attribute into one cluster CL.

  In the illustrated example, the entity e1 and the entity e7 are grouped in the cluster CL1, the entity e2 and the entity e8 are grouped in the cluster CL2, and the entity e3 and other entities are grouped in the cluster CL3. ing. Thus, an entity belonging to the cluster CL1 is obtained from a certain website A, and an entity belonging to the cluster CL2 is obtained from a website B different from the website A, and is an entity belonging to the cluster CL3. Are classified as having no URL attribute.

  Next, the determination unit 114 changes the determination condition referred to in the determination process from the condition (1) to the condition (2), and in each classified cluster, the attribute indicating the Japanese title of the web page of each entity is changed. Reference is made to determine whether or not the character strings described as the attributes of the Japanese titles of all entities included in the same cluster match each other. In the illustrated example, the character strings described as the attributes of the Japanese titles of the entities e1 and e7 included in the cluster CL1 match each other, and are described as the attributes of the Japanese titles of the entities e2 and e8 included in the cluster CL2. The character strings described do not match each other, and the character strings described as attributes of the Japanese titles of the entity e3 and other entities included in the cluster CL3 match each other. In such a case, the determination unit 114 classifies at least the cluster CL1 and the cluster CL3 as the processing target cluster CL (A) (the process of S102 in the figure), and at least the cluster CL2 as the processing non-target cluster CL (B). Sort (S104 in the figure). The processing target cluster CL (A) is a cluster in which entities that satisfy all the determination conditions up to the present time are classified, and the processing non-target cluster CL (B) satisfies one or more determination conditions up to the present time. This is the cluster into which the missing entities are classified. It is assumed that the non-processing target cluster CL (B) is not used for the subsequent processing.

  Next, the determination unit 114 determines whether or not the Japanese titles of the web pages match between the clusters included in the processing target cluster CL (A). For example, the determination unit 114 includes a character string described as the attribute of the Japanese title of the entity included in the cluster CL1 among the clusters included in the processing target cluster CL (A) and the Japanese language of the entity included in the cluster CL3. When the character strings described as title attributes are compared, and these character strings match, it is determined that the Japanese titles of the web pages match between the clusters included in the processing target cluster CL (A). To do. On the other hand, when the character strings described as the attribute of the Japanese title of the entity included in each cluster do not match, the determination unit 114 determines whether the web page Japan is between the clusters included in the processing target cluster CL (A). It is determined that the word titles do not match.

  When the Japanese title of the web page matches between the clusters included in the processing target cluster CL (A), the determination unit 114 determines the processing target cluster CL (A) as a determination target of the next determination condition ( Processing in S106 in the figure). On the other hand, when the Japanese title of the web page does not match between the clusters included in the processing target cluster CL (A), the determination unit 114 determines each cluster included in the processing target cluster CL (A) as a non-processing target. Classify into cluster CL (B). That is, the determination unit 114 does not determine each cluster included in the processing target cluster CL (A) as a determination target of the next determination condition.

  Next, the determination unit 114 changes the determination condition referred to in the determination process from the condition (2) to the condition (3), and in each cluster included in the processing target cluster CL (A) determined as the determination target, Whether or not character strings described as attributes of English titles of all entities included in the same target cluster (for example, CL1 or CL3) match each other with reference to the attribute indicating the English title of the web page of each entity Determine whether. The determination of whether or not the attributes of the English title match is the same as the determination of whether or not the attributes of the Japanese title match, and will not be described here.

  When the character strings described as the attributes of the English title match each other for all entities included in the same target cluster, the determination unit 114 classifies the cluster as the processing target cluster CL (A), Otherwise, it is classified into the processing non-target cluster CL (B).

  Next, as in the case of the Japanese title, the determination unit 114 determines whether or not the English titles of the web pages match between the clusters included in the processing target cluster CL (A). When the English titles of the web pages match between the clusters included in the processing target cluster CL (A), the determination unit 114 determines the processing target cluster CL (A) as a determination target of the next determination condition, and performs processing If the English titles of the web pages do not match between the clusters included in the target cluster CL (A), the clusters included in the processing target cluster CL (A) are classified as non-processing target clusters CL (B).

  Next, the determination unit 114 changes the determination condition referred to in the determination process from the condition (3) to the condition (4), and in each cluster included in the processing target cluster CL (A) determined as the determination target, With reference to the attribute indicating the position information and name of each entity, the character strings described as the position information attributes of all entities included in the same target cluster, and the character strings described as the name attributes Are determined to match each other.

  The determination unit 114 includes characters described as position information attributes for all entities included in the same target cluster among one or more clusters included in the processing target cluster CL (A) determined as a determination target. If the strings match and the character strings described as name attributes match, this cluster is classified as a processing target cluster CL (A), otherwise it is classified as a non-processing target cluster CL (B). To do. For example, when the position information is coordinates expressed by longitude and latitude (for example, latitude: 5.679287, longitude: 139.737386), the numerical values indicating the coordinates match up to the second decimal place, and the name of the entity When all the described character strings match, the determination unit 114 classifies the target cluster as the processing target cluster CL (A).

  As described above, after determining whether or not each entity satisfies the determination condition within the cluster and between the clusters, the determination unit 114 changes the determination condition referred to in the determination process to a new condition and repeats the determination process. . The determination unit 114 determines that the entities classified as the processing target cluster CL (A) are the same entity when the determination condition is continuously satisfied a predetermined number of times or more in the course of the repeated processing. To do. For example, as a result of the iterative process, the determination unit 114 determines that the entities classified into the processing target cluster CL (A) after satisfying all the determination conditions are the same entity. In the illustrated example, the cluster CLX and the cluster CLY satisfy all four determination conditions. Therefore, the entities e1, e7, e4 included in the cluster CLX are determined to be the same entity, and the entities e3, e9 included in the cluster CLY are determined to be the same entity. The attributes of other entities determined to be the same entity are added to the entities determined to be the same entity by the processing of the knowledge base generation unit 116 described later.

  Note that the above-described determination condition may include, for example, that the attributes indicating the securities code match between the entities to be compared with each other, or the attribute indicating the personal information (for example, the date of birth) of the person. May match each other.

  The knowledge base generation unit 116 generates a knowledge base KB based on the content CT acquired by the acquisition unit 112 and stores the knowledge base KB in the storage unit 130. For example, the knowledge base generation unit 116 refers to an HTML document included in the content CT, extracts predetermined information from the HTML document, and based on the extracted information, selects the reference content CT in the knowledge base KB. Decide whether to associate with the entity.

  For example, in an HTML document, data called structured data indicating what kind of property or attribute the content CT is may be described with a predetermined specification (for example, a specification based on schema.org). Structured data is data for allowing a web server or other crawler that provides a search site to understand information described in HTML, that is, what entity a web page should be handled by. When structured data is included in the HTML document, the knowledge base generation unit 116 acquires the structured data from the HTML document, thereby identifying the entity corresponding to the content CT in the knowledge base KB. In addition, when the HTML document does not include structured data, the knowledge base generation unit 116 specifies an entity corresponding to the content CT using a statistical probability method from the description content of the HTML document, for example. It's okay.

  And the knowledge base production | generation part 116 produces | generates the knowledge base KB by associating content CT with the entity specified based on structured data etc.

  FIG. 6 is a diagram illustrating an example of the knowledge base KB. In the figure, e_A to e_I represent entities, respectively, and arrows connecting them represent attributes (properties). As described above, the knowledge base KB is represented by a directed graph in which the above-described entity is represented as a node, the above-described attribute (property) is provided as a label, and the edge is represented as a directional edge. The directionality of the edge (the direction of the arrow in the figure) represents the upper-lower relationship of the above-described class. Although not shown, it is assumed that content CT is associated with these entities.

  In addition, when the determination unit 114 determines that the entities that are attribute comparison targets are the same entity, the knowledge base generation unit 116 integrates the comparison target entities and registers them in the knowledge base KB. “Integrating entities” means, for example, adding an attribute of the other entity to one of the comparison target entities.

  FIG. 7 is a diagram for explaining an entity integration method. For example, among the attributes of an entity e_A in the first database DB1, the name attribute is associated with the character string “XX garden”, and the latitude attribute is the character string “35.732”. Are associated with each other, the character string “139.746” is associated with the longitude attribute, and the character string “no holidays” is associated with the fixed holiday attribute. Among the attributes of a certain entity e_B in the second database DB2, the name attribute is associated with the character string “XX garden”, and the latitude attribute is the character string “35.730”. Is associated with the attribute of longitude, and the character string “139.749” is associated with the attribute of the nearest station, and the character string “□□ station” is associated with the attribute of the nearest station. When the attributes of the entity e_A and the entity e_B are compared, the character strings described as the name attributes all match, and the character strings described as the latitude and longitude attributes match up to the second decimal place. The determination unit 114 determines that these entities are the same entity. In this case, the knowledge base generation unit 116 adds the attribute of the entity e_B to, for example, the entity e_A out of the two comparison target entities, and registers it in the knowledge base KB. As a result, the entity e_A is newly added with the attribute of the nearest station that was not originally possessed. Note that the knowledge base generation unit 116 may add the attribute of the entity e_A to the entity e_B instead of the entity e_A.

  The database generation device 100 may provide the information base device 10 with the knowledge base KB generated in this way. As a result, the information providing apparatus 10 identifies an entity having a name that matches the query input in the search site or the like in the knowledge base KB, and the content CT (for example, image data) associated with the entity is identified by the query. In addition to the search result, the information is provided to a user terminal device (for example, a smartphone).

[Processing flow]
FIG. 8 is a flowchart illustrating an example of processing executed by the control unit 110 in the first embodiment. First, the acquisition unit 112 uses the communication unit 102 to communicate with the information providing apparatus 10 that provides a website and other web servers, and acquires databases managed by these apparatuses (S200).

  Next, the determination unit 114 refers to a plurality of different databases acquired by the acquisition unit 112, extracts information on a predetermined column from the information registered in each database as an entity, and extracts the information from each database. It is determined whether or not the extracted comparison target entities satisfy a predetermined determination condition (S202).

  When determining that the comparison target entities satisfy the determination condition, the determination unit 114 classifies the comparison target entities into the processing target cluster CL (A) (S204), and the comparison target entities satisfy the determination condition. If it is determined that there is no such object, the comparison target entity is classified into the processing non-target cluster CL (B) (S206).

  Next, the determination unit 114 determines whether or not determination processing has been attempted for all predetermined determination conditions (S208), and if it is determined that determination processing has not been attempted for all determination conditions, The conditions are changed (S210), and the process returns to S202 described above. In this case, the determination unit 114 repeats the process of determining entities classified into the processing target cluster CL (A) under the changed determination condition.

  On the other hand, when it is determined by the determination unit 114 that the determination process is attempted for all the determination conditions (in the above-described example, when all the four determination processes are performed), the knowledge base generation unit 116 In the processing target cluster CL (A), entities included in the same cluster are integrated and registered in the knowledge base KB (S212). Thereby, the processing of this flowchart is completed.

  According to the first embodiment described above, the attributes of entities registered in the first database DB1 are compared with the attributes of entities registered in a second database DB2 different from the first database DB1, and these entities are compared. In order to generate a database in which a plurality of comparison target entities are integrated into one entity, for example, in order to generate a database in which a plurality of comparison target entities are determined to be the same entity, It can be integrated into one entity even if the representation method of entities (for example, the types and number of attributes of entities) is different in different databases, such as a location database and a finance database. High knowledge base It can generate a KB (an example of a database).

Second Embodiment
Hereinafter, a second embodiment will be described. In the second embodiment, the first described above is that a new entity is associated with a target entity based on a character string described as an attribute of an entity to be processed (hereinafter referred to as a target entity). It is different from the embodiment. “Associating a new entity” refers to connecting a new entity as a node to an edge connected to the node when the target entity is a node in the directed graph of the knowledge base KB. Hereinafter, differences from the first embodiment will be mainly described, and description of points common to the first embodiment will be omitted. In the description of the second embodiment, the same parts as those in the first embodiment will be described with the same reference numerals.

  The determination unit 114 in the second embodiment extracts a target entity from the first database DB1 or the second database DB2, and identifies attributes that are not yet associated with other entities among the attributes of the extracted target entity. Assume that the character string representing the attribute represents the name of a candidate for an entity.

  The determination unit 114 is an entity registered in the first database DB1 or the second database DB2 or an entity that is already registered in the knowledge base KB and has similar properties to the entity candidates (hereinafter referred to as similar entities). ) Exists. The similar entity is an example of “similar object”.

  For example, the determination unit 114 determines whether or not a class to which a certain entity to be compared belongs and a class to which a candidate entity belongs belong to the same class among entities included in the database, When the classes are determined to be the same class, it is determined that the entity to be compared is a similar entity having similar properties to the entity candidates. That is, the determination unit 114 determines that a similar entity exists on the database.

  In addition, the determination unit 114 has a higher-order relationship between the class assumed to be the entity candidate and the class to which the comparison target entity belongs, and a part of the property included in the class to which the entity candidate belongs; If some of the properties included in the class to which the comparison target entity belongs are the same, the comparison target entity may be determined to be a similar entity.

  If the determination unit 114 determines that a similar entity exists, the determination unit 114 further determines whether or not an attribute described by a character string that matches the entity candidate name exists in one or more attributes of the similar entity. judge.

  The knowledge base generation unit 116 according to the second embodiment is determined by the determination unit 114 that an attribute described by a character string that matches the name of the entity candidate exists in one or more attributes of the similar entity. In this case, the entity of interest associated with the similar entity is registered in the knowledge base NB. When the database from which the similar entities are extracted is the knowledge base NB, the knowledge base generation unit 116 may associate the target entity with the similar entities already registered in the knowledge base NB.

  FIG. 9 is a diagram schematically illustrating the processing content when associating a similar entity with a target entity. For example, in the first database DB, among the attributes of the entity e_X, the name “Y Corporation” is associated with the name attribute, and the “search Taro” character string is associated with the president attribute. It is attached. In the second database DB, among the attributes of the entity e_Y, the attribute “name” is associated with the character string “search Taro”.

  In such a case, for example, the determination unit 114 sets the entity e_X as the target entity, and the character string “search Taro” described as the president attribute among the attributes of the entity e_X is used as the entity candidate name. To do. Further, the determination unit 114 sets the entity e_Y as a comparison target entity when searching for a similar entity from the database.

  The determination unit 114 compares the class to which the entity e_Y as the comparison target belongs and the class to which the candidate entity having the attribute named “Search Taro” belongs, and these classes are the same class. It is determined whether or not.

  For example, the attribute “president” is an attribute for classifying an entity having the attribute into a class “person”, and the attribute “name” is also an attribute for classifying an entity having the attribute into a class “person”. Therefore, when the character string “Search Taro” associated with the president attribute is regarded as a provisional entity, the provisional entity belongs to the class “person” and has the name attribute. It becomes the same class as the entity e_Y. Therefore, the determination unit 114 determines that the class to which the entity e_Y to be compared belongs and the class to which the candidate entity belongs are the same class. Accordingly, the determination unit 114 determines that the entity e_Y that is the comparison target is a similar entity that is similar in nature to the entity candidate obtained from the entity e_X that is the target entity.

  Then, the determination unit 114 determines whether or not an attribute described by the character string “Search Taro” as the entity candidate name exists in one or more attributes of the entity e_Y that is a similar entity. To do. In the illustrated example, the character string “Search Taro” described as the attribute of the name of the entity e_Y matches the “Search Taro” that is the candidate name of the entity.

  In such a case, the knowledge base generation unit 116 adds an entity e_X that is an entity candidate similar entity to an entity e_X that has an attribute represented by a character string that is a candidate entity name among the attributes of the entity of interest. Are registered in the knowledge base NB. As a result, among the attributes of entity e_X, the entity e_Y having the attribute of the name associated with the character string “search Taro” is also included in the attribute of president associated with the character string “search taro”. Will be associated.

  FIG. 10 is a flowchart illustrating an example of processing executed by the control unit 110 in the second embodiment. First, the acquisition unit 112 uses the communication unit 102 to communicate with the information providing apparatus 10 that provides a website and other web servers, and acquires databases managed by these apparatuses (S300).

  Next, the determination unit 114 extracts a target entity from the first database DB1 or the second database DB2, and among the attributes of the extracted target entity, a character string representing an attribute that is not yet associated with another entity, A candidate name of a certain entity is determined (S302).

  Next, the determination unit 114 includes similar entities that are similar in nature to the entity candidates among the entities registered in the first database DB1 or the second database DB2 or entities already registered in the knowledge base KB. It is determined whether or not (S304).

  If it is determined that there is no similar entity, the determination unit 114 ends the process of this flowchart. On the other hand, when it is determined that a similar entity exists, the determination unit 114 determines whether the entity candidate name matches the character string representing the attribute of the similar entity (S306).

  If it is determined that the name of the entity candidate does not match the character string representing the attribute of the similar entity, the determination unit 114 ends the process of this flowchart. On the other hand, when the determination unit 114 determines that the candidate entity name matches the character string representing the attribute of the similar entity, the knowledge base generation unit 116 associates the similar entity and the target entity with each other, Register in the knowledge base NB (S308). Thereby, the processing of this flowchart is completed.

  According to the second embodiment described above, when the character string described as the attribute of the target entity is the name of a candidate for a certain entity, the expected property of the candidate for the entity is included in the other entities. If it is determined whether or not there is a similar entity, and it is determined that there is a similar entity, a knowledge base NB that associates the similar entity with the target entity is generated. A high database can be generated.

<Hardware configuration>
The information providing apparatus 10 and the database generation apparatus 100 included in the information providing system of the above-described embodiment are realized by a hardware configuration as illustrated in FIG. 11, for example. FIG. 11 is a diagram illustrating an example of a hardware configuration of the information providing apparatus 10 and the database generation apparatus 100 according to the embodiment.

  The information providing apparatus 10 includes an NIC 10-1, a CPU 10-2, a RAM 10-3, a ROM 10-4, a secondary storage device 10-5 such as a flash memory and an HDD, and a drive device 10-6. Are connected to each other. The drive device 10-6 is loaded with a portable storage medium such as an optical disk. A program stored in a portable storage medium attached to the secondary storage device 10-5 or the drive device 10-6 is expanded in the RAM 10-3 by a DMA controller (not shown) or the like and executed by the CPU 10-2. Thus, each function is realized. The program referred to by the CPU 10-2 may be downloaded from another device via the network NW.

  The database generation device 100 includes an NIC 100-1, a CPU 100-2, a RAM 100-3, a ROM 100-4, a secondary storage device 100-5 such as a flash memory and an HDD, and a drive device 100-6. Are connected to each other. The drive device 100-6 is loaded with a portable storage medium such as an optical disk. A program stored in a portable storage medium attached to the secondary storage device 100-5 or the drive device 100-6 is expanded in the RAM 100-3 by a DMA controller (not shown) or the like and executed by the CPU 100-2. Thus, the control unit 110 is realized. The program referred to by the control unit 110 may be downloaded from another device via the network NW.

  As mentioned above, although the form for implementing this invention was demonstrated using embodiment, this invention is not limited to such embodiment at all, In the range which does not deviate from the summary of this invention, various deformation | transformation and substitution Can be added.

  DESCRIPTION OF SYMBOLS 1 ... Information provision system, 10 ... Information provision apparatus, 100 ... Database production | generation apparatus, 102 ... Communication part, 110 ... Control part, 112 ... Acquisition part, 114 ... Determination part, 116 ... Knowledge base production | generation part, 130 ... Memory | storage part

Claims (12)

The first target thing and the second target are compared with the attribute of the first target thing registered in the first database and the attribute of the second target thing registered in a second database different from the first database. A determination unit that determines whether or not the thing is the same target thing;
When the determination unit determines that the first target thing and the second target thing are the same target thing, a generation unit that generates a database that integrates the first target thing and the second target thing When,
A database generation device comprising: Each attribute of the first target thing and the second target thing is described by a character string including one or more characters in a database in which each target thing is registered,
The determination unit determines that the first target thing and the second target thing are the same target thing when satisfying a condition including matching of character strings described as attributes of each target thing. ,
The database generation device according to claim 1. The character string described as the attribute of the target thing includes a character string indicating the identification information of the resource associated with the target thing registered in the database,
The condition includes matching of character strings indicating identification information of the resource associated with the target thing,
The database generation device according to claim 2. The character string described as the attribute of the target thing includes the character string included in the content provided by the resource,
The condition includes matching of character strings included in content provided by the resource associated with the target thing,
The database generation device according to claim 3. The character string included in the content provided by the resource includes the title of the content.
The database generation device according to claim 4. The character string included in the content provided by the resource includes a character string indicating the position information of the building represented by the content.
The database generation device according to claim 4 or 5. The character string included in the content provided by the resource includes a character string indicating personal information of a person represented by the content.
The database generation device according to any one of claims 4 to 6. The character string described as the attribute of the target thing includes a character string indicating the name of the target thing,
The condition includes that the character strings indicating the names of the target objects are matched.
The database generation device according to any one of claims 2 to 7. The determination unit
While changing the condition for determining that the first target thing and the second target thing are the same target thing, the determination process is repeated,
When the condition changed in the process of repeating the determination process is continuously satisfied a predetermined number of times or more, it is determined that the first target thing and the second target thing are the same target thing.
The database generation device according to any one of claims 1 to 8. The determination unit further includes:
When the character string described as the attribute of the target thing of interest is the name of a candidate for a certain target thing, the similar target thing similar to the assumed property of the candidate for the target thing among other target things Whether or not exists,
When the determination unit determines that the similar target thing exists, the generation unit generates a database in which the similar target thing is associated with the target target object.
The database generation device according to any one of claims 2 to 9. Computer
The first target thing and the second target are compared with the attribute of the first target thing registered in the first database and the attribute of the second target thing registered in a second database different from the first database. Determine if the thing is the same subject matter,
When it is determined that the first target thing and the second target thing are the same target thing, a database integrating the first target thing and the second target thing is generated.
Database generation method. On the computer,
The first subject matter and the second subject are compared by comparing the attribute of the first subject matter registered in the first database with the attribute of the second subject matter registered in a second database different from the first database. To determine if the thing is the same subject matter,
When it is determined that the first target thing and the second target thing are the same target thing, a database integrating the first target thing and the second target thing is generated.
program.

Images