JP7375657B2 - Search program, search method, and search device - Google Patents

Description

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

A knowledge graph (KG) is known as an example of a knowledge base (KB) that collects information from various information sources.

The KG may be used, for example, in a faceted search that targets the entire data stored in the KG. Facet search is a search method that allows the user to narrow down content within a KG by selecting search conditions prepared by the data search system.

Special Publication No. 2011-513819 Special Publication No. 2005-514673

Takahiro Komamizu, Toshiyuki Amagasa, Hiroyuki Kitagawa, “D-022 Automation of facet extraction for facet search on XML data”, 13th Information Science and Technology Forum (FIT2014), Volume 2, pp. 133-134, 2014 Takahiro Komamizu, Toshiyuki Amagasa, Hiroyuki Kitagawa, “Proposal of FoX framework for faceted navigation for XML data”, 1st Forum on Data Engineering and Information Management (DEIM), B7-6, 2009

In a facet search, when searching KG data by narrowing it down to a certain specific category, the narrowing down using an appropriate key (facet key) may not be performed.

For example, in the case of a category of professional baseball player, by narrowing down the search using an appropriate facet key that is highly related to the category, such as batting appearance, handedness, and Koshien participation experience, it is possible to perform a facet search based on the user's knowledge.

However, with conventional facet search systems, even if the search is narrowed down to a specific category, the search may be narrowed down by unimportant or inappropriate facet keys such as date of birth, place of birth, company type, etc. It may be done.

As described above, if narrowing down using an appropriate facet key is not performed, there is a possibility that the number of steps it takes for the user to arrive at the desired data may increase in the facet search.

In one aspect, one object of the present invention is to output a plurality of appropriate facets from a set of facets associated with a category.

In one aspect, the search program may cause the computer to perform the following processing. The process may accept designation of a specific category. Further, the processing includes a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated with the specific category in the knowledge graph; The plurality of facets may be sequentially arranged and output according to a priority of the plurality of facets calculated based on at least one of a second index based on a distance to each of the plurality of facets.

In one aspect, the present invention can output appropriate facets from a collection of facets associated with a category.

FIG. 2 is a diagram showing an example of representation in a graph format, which is an example of a description method of RDF (Resource Description Framework). FIG. 3 is a diagram illustrating narrowing targets based on categories and facets for each category in KG data. It is a figure showing an example of a screen display of UI (User Interface) of a facet search system. FIG. 1 is a block diagram showing an example of a functional configuration of a facet search system according to an embodiment. FIG. 2 is a block diagram showing an example of a hardware configuration of a computer that implements the functions of a server. FIG. 7 is a diagram illustrating an example of frequency table creation processing performed by a frequency table creation unit. It is a figure which shows an example of a class set, a P frequency table, and a PO frequency table. It is a figure which shows an example of the calculation formula of each index. It is a figure showing an example of an index value. FIG. 2 is a diagram illustrating an example of a category importance table in which category importance is expressed in a table format. FIG. 3 is a diagram showing an example of a facet score table in which facet scores are expressed in a table format. It is a figure which shows an example of the graph showing a class and a facet key in RDF schema. It is a figure which shows an example of the graph showing a class and a facet key in RDF schema. It is a figure which illustrates a score (org_score), a facet score (new_score), a score (ont_score), and a final score (final_score) in a table format. FIG. 7 is a diagram showing an example of facet keys displayed in an item list area. FIG. 7 is a diagram showing a comparative example of MRR (Mean Reciprocal Rank) when facet keys are sorted based on “org_score” and “new_score” respectively. FIG. 7 is a diagram showing a comparative example of MRR when facet keys are sorted based on each of “new_score” and “final_score”. 3 is a flowchart illustrating an operation example of DB (Database) creation processing according to an embodiment. 19 is a flowchart illustrating an operation example of frequency table creation processing in step S1 of FIG. 18. FIG. 19 is a flowchart illustrating an example of the operation of the score DB creation process in step S2 of FIG. 18. FIG. 7 is a flowchart illustrating an operational example of facet search processing according to an embodiment.

Embodiments of the present invention will be described below with reference to the drawings. However, the embodiments described below are merely examples, and there is no intention to exclude the application of various modifications or techniques not specified below. For example, this embodiment can be modified and implemented in various ways without departing from the spirit thereof. In the drawings used in the following description, parts with the same reference numerals represent the same or similar parts unless otherwise specified.

[1] One Embodiment [1-1] Description of facet search system First, the facet search system will be briefly explained. The facet search system is, for example, a system for performing facet search on a large-scale knowledge graph (KG).

For example, the facet search system according to one embodiment improves efficiency by narrowing down candidates by category and then performing a facet search within the category, instead of performing a facet search on the entire search target data in the KG. Realize.

The data stored in the KG is expressed using a description method called RDF (Resource Description Framework), which has a set of three elements: a subject, a predicate, and an object.

In faceted search using KG, a category is a class of instances. Further, the facet key becomes a "predicate" and the facet value becomes an "object". Through facet search, the facet search system searches for a set of instances, such as "subjects," that match the search conditions.

Here, an "instance" represents an event or thing in the world, and is also called an entity. For example, the baseball player "Taro Yamada" is an example of an instance, and may be represented by a character string such as "Taro Yamada" or by an ID in a DB (Database).

Examples of DBs include various DBs that store data (open data) and publish it on the Internet, such as DBs that use LOD (Linked Open Data) technology.

The ID in the DB is the URI (Uniform Resource Indicator) of the location where information on "Taro Yamada" can be referenced in the DB, for example, the URL (Uniform Resource Locator) of the web page containing the article on "Taro Yamada". can be mentioned. As an example, if the domain of the DB is "aaa.org", the ID of "Taro Yamada" in the DB is "http://aaa.org/resource/Taro_Yamada".

In one embodiment, IDs in one or more DBs may be settable as instances in the KG. In other words, the KG is a knowledge base that collects information from one or more DBs as information sources.

"Class" represents the type of instance; for example, the class of "Taro Yamada" is "baseball player." Instances may belong to multiple classes; “Taro Yamada” is a “baseball player” “http://aaa.org/ontology/BaseballPlayer” and “athlete” “http://aaa.org/ ontology/Athlete” and “person” “http://aaa.org/ontology/Person”.

Classes can maintain superior and subordinate relationships. For example, when using the RDF schema, “BaseballPlayer” and “Athlete” have a relationship of “rdfs:subClassOf” (subclass).

Hereinafter, RDF Schema (which may also be referred to as RDFS Schema), and schemas defined using other standard vocabularies or uniquely defined vocabularies will be referred to as "ontologies."

FIG. 1 is a diagram showing an example of representation in a graph format, which is an example of an RDF description method. In the explanation of FIG. 1, an example will be shown in which the URI in the DB "http://aaa.org/" is included in the instance for "Hanako Suzuki", the governor of Tokyo.

Further, the expression example shown in FIG. 1 is shown below in text (n3) format.

<http://ja.aaa.org/resource/Tokyo><http://aaaa.org/ontology/leader><http://ja.aaa.org/resource/HanakoSuzuki>.
<http://ja.aaa.org/resource/HanakoSuzuki>
rdf:type <http://aaa.org/ontology/Politician>;
<http://aaa.org/ontology/birthPlace><http://ja.aaa.org/resource/HyogoPrefecture>;
<http://aaa.org/ontology/birthDate> “1960-01-01”.

In this way, RDF expresses "things" and "things" using the triplet of "subject" (S), "predicate" (P), and "object" (O). . According to the expression examples shown in FIG. 1 and the above text (n3) format, "things" and "things" are organized as follows. Note that "rdf:type" is a predicate that defines the relationship between an instance and a class.

“The governor (P) of Tokyo (S) is Hanako Suzuki (O).”
"Hanako Suzuki (S) is (P) a politician (O), her hometown (P) is Hyogo Prefecture (O), and her date of birth (P) is 1960-01-01 (O)."

FIG. 2 is a diagram illustrating narrowing targets based on categories and facets for each category in KG data. For example, the KG data includes a set of "Taro Yamada", "rdf:type", and "professional baseball player" as a set of "subject", "predicate", and "object", and a set of "Taro Yamada" and "turn at bat". , may include the "left" set.

The facet search system can be used, for example, from "things" to "people" or "organizations," from "people" to "politicians," "professional baseball players" or "soccer players," from "organizations" to "companies," or "professional baseball players." It may be possible to narrow down (search) categories hierarchically, such as "Team".

For example, in a facet search system, a user can narrow down search results by selecting a facet in any category and selecting a value that is a presented candidate. In addition, in the example of Figure 2, if we focus on the category "politician", the facets are {name, political party affiliation, place of birth, date of birth}, and the character strings and IDs in the DB that are associated with each are This is the value for the facet.

FIG. 3 is a diagram showing an example of a screen display of a UI (User Interface) of the facet search system. As shown in FIG. 3, the UI of the facet search system may include a category search screen 100 and a facet search screen 200 transitioned from the category search screen 100 as screens displayed on the user's terminal.

The category search screen 100 is a screen for narrowing down categories from KG data, in other words, a screen for accepting the specification of a specific category, and may include, for example, a category selection area 110 and a class selection area 120. Note that the category search screen 100 may include a button 130 for displaying search conditions saved on the facet search screen 200.

The selection area 110 is, for example, an area for accepting the selection of a dataset, such as a ``person'' or an ``organization,'' from the ``things'' shown in FIG. 2, and in the example of FIG. B" is displayed. The selection area 120 is an area that displays a list of classes associated with the "person" or "organization" selected in the selection area 110, and includes, for example, "person" to "politician" shown in FIG. This is an area for accepting selections such as "professional baseball player" or "soccer player."

In the example of FIG. 3, for convenience, the selection area 110 is referred to as a "dataset" and the selection area 120 is referred to as a "class" selection area; A "class" can be thought of as a "category."

The UI displays a facet search screen 200 based on information on the class (category) selected in the selection area 120.

The facet search screen 200 is a screen for searching for facets from the class (category) selected on the category search screen 100. That is, the facet search screen 200 is a search screen that is a transition destination from the category search screen 100, and is a search screen for performing a search targeting a plurality of facets associated with a specific category.

The facet search screen 200 may include, for example, an item list area 210, a search condition setting area 230, an output item setting area 250, an output language setting area 260, and a list display area 280.

The item list area 210 is an area that displays facet keys in the KG data based on the class (category) selected on the category search screen 100.

The search condition setting area 230 is an area for setting search conditions for the facet key selected in the item list area 210 and the add button 220 pressed. The output item setting area 250 is an area for setting an item for outputting an entity and an output order for a facet key that is selected in the item list area 210 and the add button 240 is pressed. The output language setting area 250 is an area for setting the output language of the entity.

The list display area 280 is an area that displays a list of entities based on the settings in the setting areas 230, 250, and 260 when the search button 270 is pressed.

Note that the facet search screen 200 displays operation buttons 290 for any one or more of the display contents of the list display area 280, the setting contents of the setting areas 230, 250, and 260, and the display contents of the item list area 210. It's okay. The operation buttons 290 include, for example, an output button in CSV (Comma Separated Value) format, a button to confirm a SPARQL (SPARQL Protocol and RDF Query Language) sentence, which is an example of an RDF query language, a button to save search conditions, etc. may be included.

Incidentally, the item list area 210 on the facet search screen 200 may not display facet keys suitable for a category. Hereinafter, such a case will be explained as a comparative example.

(Comparative example)
For example, assume that the screen display shown in FIG. 3 is executed using the technique described in Non-Patent Document 1. In this technique, for example, the server uses XML (eXtensible Markup Language) data to extract facet values in the following procedure.

(i) The server extracts the structure summary from the XML data.
The structure summary is information indicating the parent-child relationship of each element in the XML data, the appearance frequency of child elements with respect to the parent element, and the like.

(ii) Extract class candidates and facet candidates from the structure summary.
In the above technique, a certain node in the structure summary that corresponds to an object in the XML data is called a class, and a selected node among the descendant nodes of the class node in the structure summary is called a facet.

(iii) Extract appropriate ones from class candidates and facet candidates.
(iv) Extract the XML subtree corresponding to the class extracted from the XML data as an object.
(v) The value of the element corresponding to the extracted facet is extracted as the value of the facet.

For example, in steps (i) to (v) above, the server may target KG data in RDF description format instead of XML data.

In the technique described in Non-Patent Document 1, in the procedure (iii) above, the server employs a method that combines a frequency-based approach and a meaning-based approach. In the frequency approach, the server can extract facet keys that can obtain more instances in the KG data. In a semantic-based approach, the server uses existing knowledge, such as WordNet or Wikipedia, to determine whether a facet key is human-interpretable.

Note that the above technique does not take into account the importance of facet keys. On the other hand, Non-Patent Document 2 describes an index for ranking facet keys.

Here, consider a case where the item list area 210 on the facet search screen 200 is displayed using the techniques described in Non-Patent Documents 1 and 2, in other words, a case where a facet search within a certain specific category is performed. . When performing a facet search within a certain specific category, there are cases where it is better to narrow down the search using a facet key unique to the category. For example, assume that the category "professional baseball player" is selected on the category search screen 100.

In this case, with the technique described in Non-Patent Document 2, due to the statistical tendency of the entire KG data, hometown and date of birth are ranked higher as important facet keys than Koshien participation experience.

Further, the technique described in Non-Patent Document 1 only uses whether or not a facet exists in a knowledge base, and does not use the ontology of knowledge. For example, for a facet key of "date," a more important facet would be "date of birth" for a person, and "date of first appearance" for a professional baseball player. This information is structured in the KG using ontology and the like.

However, when the category "Professional Baseball Player" is selected, it would be better to narrow down the search based on facet keys such as batting order, handedness, and Koshien participation experience, rather than date of birth, place of birth, company type, etc., based on the user's knowledge. Faceted search can be realized.

Therefore, the facet search system according to an embodiment described below mainly employs the methods (a) and (b) below to display rankings of facet keys suitable for categories in the item list area 210. Do the following.

(a) Introduction of indicators that emphasize category-specific facet keys.
(b) Implementation of semantic interpretation of facet keys considering KG's knowledge structure.

Thereby, it is possible to output a plurality of appropriate facets from a set of facets associated with a category, and it is possible to reduce the number of steps it takes for the user to arrive at the desired data.

[1-2] Configuration example of facet search system FIG. 4 is a block diagram showing a functional configuration example of the facet search system 1 according to one embodiment. The facet search system 1 is an example of a search system that performs a facet search, and as shown in FIG. 4, it exemplarily includes a server 2, a knowledge graph (KG) 3, and one or more ) terminal 4 may be provided.

KG3 is an example of a knowledge base, and may store data written in an RDF description format, for example.

The terminal 4 is an example of an information processing terminal used by a user of the facet search system 1, and is a computer such as a PC (Personal Computer) or a server that accesses the server 2 regarding facet searches.

The KG 3 and the server 2 and the terminal 4 and the server 2 may be connected to each other via a network (not shown) so that they can communicate with each other. The network may include a WAN (Wide Area Network), a LAN (Local Area Network), or a combination thereof. A WAN may include the Internet, and a LAN may include a VPN (Virtual Private Network).

The server 2 is an example of a search device, an information processing device, or a computer. For example, in the facet search system 1, the server 2 performs various processes such as referring to the KG 3, responding to the terminal 4, and notifying information in response to various accesses related to facet searches from the terminal 4.

The server 2 may, for example, provide the terminal 4 with a function to enable access. Examples of such functions include generation and display control of screens such as web pages used for access by the terminal 4. For example, the terminal 4 sends an access request to the server 2 using an application such as a browser, and accesses the server 2 via a web page displayed on the application based on screen information received from the server 2. good.

The server 2 may be a virtual server (VM; Virtual Machine) or a physical server. Further, the functions of the server 2 may be realized by one computer, or may be realized by two or more computers. Furthermore, at least some of the functions of the server 2 may be realized using HW (Hardware) resources and NW (Network) resources provided by a cloud environment.

(Hardware configuration example)
FIG. 5 is a block diagram showing an example of the hardware (HW) configuration of the computer 10 that implements the functions of the server 2. As shown in FIG. When a plurality of computers are used as HW resources to implement the functions of the server 2, each computer may have the HW configuration illustrated in FIG. 5.

As shown in FIG. 5, the computer 10 includes, as an example, a processor 10a, a memory 10b, a storage section 10c, an IF (Interface) section 10d, an I/O (Input/Output) section 10e, and a reading section as an HW configuration. 10f may be provided.

The processor 10a is an example of an arithmetic processing device that performs various controls and calculations. The processor 10a may be communicably connected to each block within the computer 10 via a bus 10i. Note that the processor 10a may be a multiprocessor including a plurality of processors, a multicore processor having a plurality of processor cores, or a configuration including a plurality of multicore processors.

Examples of the processor 10a include integrated circuits (ICs) such as a CPU, MPU, GPU, APU, DSP, ASIC, and FPGA. Note that a combination of two or more of these integrated circuits may be used as the processor 10a. CPU is an abbreviation for Central Processing Unit, and MPU is an abbreviation for Micro Processing Unit. GPU is an abbreviation for Graphics Processing Unit, and APU is an abbreviation for Accelerated Processing Unit. DSP is an abbreviation for Digital Signal Processor, ASIC is an abbreviation for Application Specific IC, and FPGA is an abbreviation for Field-Programmable Gate Array.

The memory 10b is an example of HW that stores information such as various data and programs. Examples of the memory 10b include one or both of a volatile memory such as a DRAM (Dynamic Random Access Memory), and a non-volatile memory such as a PM (Persistent Memory).

The storage unit 10c is an example of HW that stores information such as various data and programs. Examples of the storage unit 10c include various storage devices such as magnetic disk devices such as HDDs (Hard Disk Drives), semiconductor drive devices such as SSDs (Solid State Drives), and nonvolatile memories. Examples of nonvolatile memory include flash memory, SCM (Storage Class Memory), and ROM (Read Only Memory).

Furthermore, the storage unit 10c may store a program 10g (search program) that implements all or part of various functions of the computer 10. For example, the processor 10a of the server 2 can implement the functions of the server 2 illustrated in FIG. 4 by loading the program 10g stored in the storage unit 10c into the memory 10b and executing it.

The IF unit 10d is an example of a communication IF that performs connection with a network, control of communication, and the like. For example, the IF section 10d may include an adapter compliant with LAN (Local Area Network) such as Ethernet (registered trademark), optical communication such as FC (Fibre Channel), or the like. The adapter may be compatible with one or both of wireless and wired communication systems. For example, the server 2 may be communicably connected to each of the KG 3 and the terminal 4 via the IF section 10d. Further, for example, the program 10g may be downloaded from the network to the computer 10 via the communication IF and stored in the storage unit 10c.

The I/O unit 10e may include one or both of an input device and an output device. Examples of the input device include a keyboard, mouse, touch panel, and the like. Examples of the output device include a monitor, a projector, and a printer.

The reading unit 10f is an example of a reader that reads data and program information recorded on the recording medium 10h. The reading unit 10f may include a connection terminal or device to which the recording medium 10h can be connected or inserted. Examples of the reading unit 10f include a USB (Universal Serial Bus) compliant adapter, a drive device that accesses a recording disk, a card reader that accesses a flash memory such as an SD card, and the like. Note that the program 10g may be stored in the recording medium 10h, or the reading unit 10f may read the program 10g from the recording medium 10h and store it in the storage unit 10c.

Examples of the recording medium 10h include non-transitory computer-readable recording media such as magnetic/optical disks and flash memories. Examples of magnetic/optical discs include flexible discs, CDs (Compact Discs), DVDs (Digital Versatile Discs), Blu-ray discs, and HVDs (Holographic Versatile Discs). Examples of flash memory include semiconductor memories such as USB memory and SD cards.

The HW configuration of the computer 10 described above is an example. Therefore, the number of HWs within the computer 10 may be increased or decreased (eg, adding or deleting arbitrary blocks), dividing, integrating in any combination, adding or deleting buses, etc., as appropriate. For example, in the server 2, at least one of the I/O section 10e and the reading section 10f may be omitted.

Note that the terminal 4, which is an example of an information processing terminal, may be realized by the same HW configuration as the computer 10 described above.

For example, the processor 10a of the terminal 4 can implement the functions of the terminal 4 shown in FIG. 4 by loading the program 10g stored in the storage unit 10c into the memory 10b and executing it.

Note that the terminal 4 shown in FIG. 4 may include an input device and a display device, which are examples of the I/O section 10e. For example, the processor 10a of the terminal 4 may display each screen on the display device based on information received from the server 2 via the IF section 10d. Further, the processor 10a of the terminal 4 may transmit the input information to the server 2 via the IF section 10d.

(Functional configuration example)
Returning to the explanation of FIG. 4, the server 2 may include, for example, a memory section 21, a search control section 22, a statistical processing section 23, a meaning interpretation processing section 24, and a ranking adjustment section 25.

The memory unit 21 is an example of a storage area, and stores various information regarding facet searches. As shown in FIG. 4, the memory unit 21 may illustratively store a frequency table 21a and a score DB 21b. In the following description, for convenience, the data formats of the frequency table 21a and the score DB 21b will be explained as table formats, but the data formats are not limited to this and may be data formats of various DBs.

Note that the frequency table 21a and the score DB 21b may be stored in a storage area included in at least one of the memory 10b and the storage unit 10c shown in FIG. 5, for example. In other words, the memory section 21 may be realized by a storage area included in at least one of the memory 10b and the storage section 10c.

The search control unit 22 provides the terminal 4 with a UI including a category search screen 100 and a facet search screen 200 illustrated in FIG. For example, the search control unit 22 generates screen information for each of the category search screen 100 and the facet search screen 200, transmits it to the terminal 4, and indicates character strings and selection items input via the UI on the terminal 4. Control information may be received from the terminal 4.

As an example, when the search control unit 22 receives control information including the classes (categories) selected in the selection areas 110 and 120 of the category search screen 100 from the terminal 4, the search control unit 22 sends the control information to the statistical processing unit 23 and the semantic interpretation process. Each of the sections 24 may be notified. In other words, the search control unit 22 is an example of a reception unit that accepts designation of a specific category.

Further, when the search control unit 22 is notified of information indicating a plurality of facet keys from the ranking adjustment unit 25, the search control unit 22 displays the facet keys included in the information, including their display order (arrangement order), in the item list area 210. to be displayed. In other words, the search control unit 22 is an example of an output unit that sequentially arranges and outputs a plurality of facets according to the priority calculated by the ranking adjustment unit 25 (for example, displays them in the item list area 210).

The statistical processing unit 23 creates or updates the score DB 21b based on the KG3 and control information as a preliminary phase (preparation phase) before the facet search is performed. For example, the statistical processing unit 23 may include a frequency table creation unit 23a and a score calculation unit 23b.

The frequency table creation unit 23a creates a frequency table 21a. For example, as shown in FIG. 6, the frequency table creation unit 23a acquires all class sets 21a1 to be subjected to facet search from the KG3. For example, the frequency table creation unit 23a may execute the class acquisition query Q1 shown in FIG. 6 to acquire the class set 21a1 from KG3.

Then, the frequency table creation unit 23a calculates a facet key frequency table 21a2 and a facet key/facet value frequency table 21a3 (see FIG. 7) based on the classes included in the class set 21a1 and KG3. For example, the frequency table creation unit 23a executes the facet key frequency table query Q2 and the facet key/facet value frequency table query Q3 shown in FIG. Table 21a3 may be obtained.

Note that in KG3, the class corresponds to the subject (S), the facet key corresponds to the predicate (P), and the facet value corresponds to the object (O). Therefore, the facet key frequency table 21a2 may be referred to as the P frequency table 21a2, and the facet key/facet value frequency table 21a3 may be referred to as the PO frequency table 21a3.

Each of the queries Q1 to Q3 shown in FIG. 6 is an example of a query using SPARQL, which is an example of an RDF query language. In queries Q1 to Q3, "?s" corresponds to the subject (S), "?p" corresponds to the predicate (P), "?o" corresponds to the object (O), and "%CLASS%" in queries Q2 and Q3. is replaced by each class obtained in query Q1.

FIG. 7 is a diagram showing an example of the class set 21a1, the P frequency table 21a2, and the PO frequency table 21a3.

As illustrated in FIG. 7, the class set 21a1 is information obtained by extracting facet keys (P) such as "name" and "political party" for each category such as "politician" from KG3.

The P frequency table 21a2 is information obtained by extracting the "frequency" in KG3 (for example, the "number" of records obtained by query Q2) for each facet key (P) included in the class set 21a1 from KG3.

The PO frequency table 21a3 shows the "frequency" in KG3 (for example, the "number of records obtained by query Q3") for each facet key (P) and each facet value (O) included in the class set 21a1 from KG3. ”) is extracted information.

The frequency table creation unit 23a may store information on at least the PO frequency table 21a3 among the class set 21a1, the P frequency table 21a2, and the PO frequency table 21a3 as the frequency table 21a in the memory unit 21. The reason why at least the information of the PO frequency table 21a3 is stored as the frequency table 21a is that the P frequency table 21a2 can be derived by summing the facet values of the PO frequency table 21a3 for each facet key.

The score calculation unit 23b calculates the score DB 21b based on the frequency table 21a created by the frequency table creation unit 23a.

FIG. 8 is a diagram showing an example of a calculation formula for each index, and FIG. 9 is a diagram showing an example of the index value 21b1. For example, the score calculation unit 23b may calculate the index value 21b1 illustrated in FIG. 9 using the calculation formulas for each index illustrated in formulas (1) to (4) below (see FIG. 8).

The facet frequency (freq(f)) shown in the above formula (1) is an index indicating the frequency of appearance of a facet in the entire search target data of KG3. A facet with a higher facet frequency means that it appears in more search target data, and a facet with a lower facet frequency means that it appears more locally in the search target data. In the above formula (1), "N" is the total number of data to be searched, and "n(facet)" is the number of data to be searched per facet.

The facet balance degree (bala(f)) shown in the above equation (2) is the distribution of the number of data that can be searched for each facet. A facet with a higher degree of facet balance has a wider range of data that can be searched and is a more balanced facet, and a facet with a lower degree of facet balance means that the occurrence of facet keys in the search target data is more biased. In the above formula (2), “n(key _i )” is the number of search target data that can be searched with the facet key “key _i ”, “n _key ” is the number of keys in the facet “f”, and “μ” is the average of “n(key _i )”.

The key density (card(f)) shown in the above equation (3) is the distribution of the number of facet keys in each facet. In the above equation (3), "n _key " is the number of keys in facet "f", "μ" is the average of "n _key ", and "σ ² " is the variance.

The key monotonicity (mono(f)) shown in the above equation (4) is an index of the monotonicity of a facet key, and is an index for evaluating the average number of data that can be retrieved for each facet key in a certain facet. In the above equation (4), "avg" is the average of "n(key _i )", and "μ" and "σ ² " are the average and variance thereof.

Note that the above formulas (1) to (4) are similar to the formulas for calculating each index described in, for example, Non-Patent Document 2, and detailed explanations thereof will be omitted.

The score calculation unit 23b may calculate the index value 21b1 for each facet key, as illustrated in FIG. 9, based on the frequency table 21a and the above equations (1) to (4).

Furthermore, the score calculation unit 23b calculates the category importance degree (signif(f, C)) for each category and each facet key based on the frequency table 21a and the index value 21b1. Category importance (signif(f, C)) is an index that takes into account the frequency of appearance of a facet within a class, and is an index that emphasizes a facet key unique to a category.

FIG. 10 is a diagram showing an example of a category importance table 21b2 that represents the category importance (signif(f, C)) in a table format.

For example, the score calculation unit 23b may calculate the category importance degree (signif(f, C)) based on the following formula (5).
signif(f, C) = weight(f, C) * uniq(f) (5)

In the above equation (5), "weight(f, C)" is the weight of the appearance frequency of the facet within the class, and may be expressed by, for example, the following equation (6).
weight(f, C) = n(f, C) / N _d (6)

Here, in the above equation (6), "N _d " is the total number of "f" in class "C", and "n(f, C)" is the number of facets in class "C".

Furthermore, in the above equation (5), "uniq(f)" is an index indicating the degree of uniqueness of each facet. Uniqueness is an index indicating whether a facet appears biasedly in a specific class (for example, only in a specific class) among a plurality of classes. For example, "uniq(f)" may be represented by the following formula (7).
uniq(f) = NC / uniq_count(f) (7)

Here, in the above equation (7), "NC" is the total number of classes in KG3, and "uniq_count(f)" is the number of classes in KG3 in which "f" is included. Thus, uniqueness may be obtained by dividing the total number of classes in KG3 by the number of classes in which the facet appears.

Thus, category importance (signif(f, C)) is an example of a first index based on the number of categories with which each of the plurality of facets associated with a particular category in KG3 is associated in KG3. .

The score calculation unit 23b may calculate the category importance degree (signif(f, C)) for each facet key using the frequency table 21a and the above equation (5). Note that, as shown in FIG. 10, the score calculation unit 23b normalizes the category importance (signif(f, C)) to a range of "0" to "1".

Then, the score calculation unit 23b calculates a facet score (new_score) using the score (org_score) based on the index value 21b1 described above and the score (signif) based on the above formula (5).

For example, the score calculation unit 23b may calculate a weighted linear sum of the score (org_score) and the score (signif), as exemplified by the following formula (8), and obtain the facet score (new_score).
new_score(f) = ω * org_score(f) + (1 - ω) * signif(f) (8)

In the above equation (8), "ω" is a weight. In one embodiment, as a non-limiting example, ω=0.5. In the case of "ω=0.5", the above equation (8) is expressed as the following equation (8').
new_score(f) = 0.5 * org_score(f) + 0.5 * signif(f) (8')

Here, in the above formula (8) or (8'), "org_score(f)" may be expressed by the following formula (9). Note that in the following formula (9), it is assumed that α+β+γ+θ=1.

FIG. 11 is a diagram showing an example of a facet score table 21b3 that represents facet scores (new_score) in a table format. For example, the score calculation unit 23b may store information of at least the facet score table 21b3 among the index value 21b1, the category importance table 21b2, and the facet score table 21b3 in the memory unit 21 as the score DB 21b.

As illustrated in Figure 11, the score (new_score) that takes into account the importance of facets within a class (category) is used to determine whether, for example, "politician" has "affiliation" rather than "name" or "date of birth." "Political party" has a higher score, and for "professional baseball player", "defensive position" has a higher score than "name" and "date of birth".

That is, in the item list area 210 shown in FIG. 3, facet keys such as "political party affiliation" for "politicians" and "defensive position" for "professional baseball players" are displayed as items with high priority. become. Therefore, the server 2 can output a plurality of appropriate facets narrowed down by facet keys with higher scores from the set of facets associated with the category.

In this way, the score calculation unit 23b calculates the category importance including uniq(f) for each of the plurality of facets, and stores information (for example, the score DB 21b) including the calculated category importance in the memory unit 21. This is an example of a calculation unit.

Returning to the explanation of FIG. 4, when a facet search is performed, the semantic interpretation processing unit 24 performs a search from the class (category) selected on the category search screen 100 (see FIG. 3) to the class to which a certain facet belongs. The pass on KG3 is reflected in the score. In this manner, the semantic interpretation processing unit 24 changes the score so that a facet with a high priority in consideration of ontology is extracted according to the selected class when a facet search is performed.

For example, the semantic interpretation processing unit 24 may calculate a score (ont_score) that takes into account the distance on the graph in the ontology, as shown in equation (10) below. The score (ont_score) is an example of a second index based on the distance between the selected class and each of the plurality of facets in KG3.
ont_score(f, C) = 1 / (distance(C, C _f ) + 1) (10)

Here, in the above formula (10), "f" is the facet for which the score is calculated, "C" is the class (category) selected in the item list area 210, and "C _f " is the item This is the class to which the candidate facets displayed in the list area 210 belong.

In addition, in the above equation (10), “distance(C, C _f )” is the distance on the schema (ontology), that is, on the graph, between the selected class and the class to which the selected facet belongs. . The distance on the graph may mean the distance between layers of classes (categories). For example, the semantic interpretation processing unit 24 may regard each hierarchical class as each node in a tree structure, and calculate the distance between nodes as a distance on a graph using a known method. Note that when “C = C _f ”, “distance(C, C _f ) = 0”.

FIG. 12 is a diagram showing an example of a graph representing classes and facet keys in the RDF schema. An example of calculating the score (ont_score) when the selected category is "BaseballPlayer" (see broken line) in the example of FIG. 12 will be described below.

As an example, when the facet for which the score is calculated is "name (person)", the score (ont_score) is calculated as shown in equation (11) below.
ont_score(name(person), BaseballPlayer)
= 1 / (distance(BaseballPlayer, Person) + 1)
= 1 / (1 + 1) = 0.5 (11)

As another example, when the facet for which the score is calculated is "headquarters location", the score (ont_score) is calculated as shown in equation (12) below.
ont_score(head office location, BaseballPlayer)
= 1 / (distance(BaseballPlayer, Company) + 1)
= 1 / (3 + 1) = 0.25 (12)

As another example, when the facet for which the score is calculated is "political party affiliation," the score (ont_score) is calculated as shown in equation (13) below.
ont_score(party affiliation, BaseballPlayer)
= 1 / (distance(BaseballPlayer, Politician) + 1)
= 1 / (2 + 1) = 0.33 (13)

In this way, the semantic interpretation processing unit 24 assigns a higher priority to facet keys with a higher degree of relevance in the ontology, for example, the closer the distance on the graph (the smaller the distance(C, C _f )). Calculate the score (ont_score). In addition, the semantic interpretation processing unit 24 uses facet keys that have a smaller degree of relevance in the ontology, for example, a facet key that is farther away on the graph (larger distance (C, C _f )), the lower the priority is. Calculate (ont_score). In other words, the score (ont_score) can be said to be a score that reflects the semantic interpretation of the facet key in consideration of the knowledge structure of KG3.

In addition to or instead of the distance on the graph (distance(C, C _f )) between the facet key on KG3 and the selected class, the semantic interpretation processing unit 24 calculates the semantic distance. A score (ont_score) may be calculated based on the considered index.

An example of an index that takes semantic distance into consideration (semantic distance index) is an index that is determined depending on whether the vocabulary of a predicate (P; Predicate) serving as a facet key is a standard vocabulary. The determination as to whether the vocabulary is a standard vocabulary may be made, for example, by determining whether the vocabulary serving as a facet key is registered in a DB that stores standard vocabulary. Examples of DBs that store standard vocabulary include DBs such as "prefix.cc" and "Linked Open Vocabularies."

For example, the semantic interpretation processing unit 24 may calculate a score (ont_score) that takes into account the semantic distance in the ontology, as shown in equation (14) below.
ont_score(f, C) = (1 / (distance(C, C _f ) + 1)) * std_vocab(f) (14)

Here, in the above equation (14), the term (1 / (distance(C, C _f ) + 1)) is the same as the above equation (10), and "std_vocab(f)" is expressed as the following equation ( As shown in 15), the function may be "1.0" if it is a standard vocabulary, "0.5" if it is not a standard vocabulary, etc.

For example, as shown in the above equation (14), the semantic distance index shown in the above equation (15) is calculated based on the score considering the distance on the graph shown in the above equation (10). By multiplying , a score may be calculated that takes into account both the graphical and semantic distances.

Alternatively, instead of the above equation (14), the semantic interpretation processing unit 24 may adopt "std_vocab(f)" shown in the above equation (15) as the score (ont_score) considering only the semantic distance. good.

FIG. 13 is a diagram showing an example of a graph representing classes and facet keys in the RDF schema. Hereinafter, in the example of FIG. 13, an example of calculating the score (ont_score) in consideration of the semantic distance index will be described when the selected category is "BaseballPlayer" (see broken line).

In Figure 13, for example, "name (person)" is a standard vocabulary expressed as "foat:name", and "name (company)" is a standard vocabulary expressed as "skos:prefLabel". Yes, and "head office location" is a unique vocabulary (non-standard vocabulary) expressed as "14a-ont:head office location".

As an example, when the facet for which the score is calculated is "name (company)", the score (ont_score) is calculated as shown in equation (16) below.
ont_score(name(company), BaseballPlayer)
= (1 / (distance(BaseballPlayer, Company) + 1)) * std_vocab(name(company))
= (1 / (3 + 1)) * 0.5 = 0.125 (16)

As another example, when the facet for which the score is calculated is "headquarters location", the score (ont_score) is calculated as shown in equation (17) below.
ont_score(head office location, BaseballPlayer)
= (1 / (distance(BaseballPlayer, Company) + 1)) * std_vocab(Head office location)
= (1 / (3 + 1)) * 1.0 = 0.25 (17)

In this way, the semantic interpretation processing unit 24 uses a score (ont_score) such that the facet key with a higher degree of relevance in the ontology, for example, the facet key with a closer (larger) semantic distance (std_vocab), has a higher priority. Calculate. In addition, the semantic interpretation processing unit 24 calculates a score (ont_score) such that the facet key with a smaller degree of relevance in the ontology, for example, a facet key with a farther (smaller) semantic distance (std_vocab), has a lower priority. do.

As a result, in the item list area 210 shown in FIG. 3, facet keys that are close to the selected category in one or both of the graph distance and semantic distance are displayed as high priority items. It turns out. Therefore, the server 2 can output a plurality of appropriate facets narrowed down by facet keys with higher scores from the set of facets associated with the category.

Returning to the explanation of FIG. 4, the ranking adjustment unit 25 calculates a final facet score (final_score) in the facet search.

For example, in a facet search, the above-mentioned search control unit 22 selects a target category while tracing the hierarchy of classes assigned to entities on the category search screen 100, and displays the facet search screen 200.

At this time, based on the selected category, the semantic interpretation processing unit 24 calculates a score (ont_score) as a facet importance level using the knowledge structure.

The ranking adjustment unit 25 calculates a final score (final_score) based on the facet score (new_score) calculated by the score calculation unit 23b in the preliminary phase and the score (out_score) calculated by the semantic interpretation processing unit 24. It's fine. For example, the ranking adjustment unit 25 may calculate the score (final_score) by multiplying the facet score (new_score) and the score (out_score) based on the following formula (18).
final_score(f, C) = new_score(f, C) * ont_score(f, C) (18)

FIG. 14 shows the score (org_score) shown in the above equation (9), the facet score (new_score) shown in the above equation (8), the score (ont_score) shown in the above equation (10) or the above equation (14), and the above equation It is a figure which illustrates the final score (final_score) shown in (18) in a table format.

The example in FIG. 14 shows a case where "professional baseball player" is selected as the category on the category search screen 100. In this case, the final score (final_score) is "0.562", which is higher for the "defense position" where both the facet score (new_score) and the score (ont_score) are high. This "defense position" has a lower score than the "name" and "date of birth" in the score (org_score) shown in equation (9) above.

In this way, by using an index that emphasizes category-specific facet keys and a semantic interpretation of facet keys that takes into account the knowledge structure of KG3, appropriate facets can be determined, such as "defensive position" for the category of "professional baseball player." Calculated so that the key score is high.

For example, the ranking adjustment unit 25 sorts the facet keys displayed in the item list area 210 in descending order of the score based on the calculated final score (final_score), and the information on the sorted facet keys is sent to the search control unit 22. You can output to

The final score (final_score) calculated by the ranking adjustment unit 25 is an example of the priorities of the plurality of facets.

As a result, as illustrated in FIG. 15, the search control unit 22 displays the final score (final_score)-based list in the item list area 210 instead of the score (org_score)-based list 211 shown in equation (9) above. A list 212 of facet keys can be displayed. The list 212 is an example of a list in which a plurality of facets are arranged in order according to the priorities calculated by the ranking adjustment unit 25.

FIGS. 16 and 17 are diagrams showing comparative examples of MRR (Mean Reciprocal Rank) when facet keys are sorted based on "org_score", "new_score", and "final_score", respectively. MRR is an evaluation index of the quality of search results, and has a value in the range of "0" to "1". The closer the MRR is to "1", the higher the facet keys displayed in the item list area 210 are searched (selected), which means that the facet search is realized in accordance with the user's knowledge. .

FIG. 16 shows a comparison example between the facet ranking and MRR based on the score (org_score) shown in the above equation (9) and the facet ranking and MRR based on the facet score (new_score) shown in the above equation (8). In the example in Figure 16, the category "Baseball Player" is selected from the dataset with the number of data to be faceted "11401", and the facet keys "At-bat", "Draft ranking", and "First appearance" are selected. Assume a case.

As illustrated in FIG. 16, based on the "org_score", the facet rankings are around 100 for all facet keys, and the MRR is "0.0094". On the other hand, based on the "new_score", the facet rankings are 3rd to 6th, and the MRR is "0.2333".

In this manner, in the "new_score" base, the facet rankings and MRR of "baseball player"-specific facets, such as "turn at bat," "draft ranking," and "first appearance," are higher than in the "org_score" base.

FIG. 17 shows a comparison example between the facet ranking and MRR based on the facet score (new_score) shown in the above equation (8) and the facet ranking and MRR based on the final score (final_score) shown in the above equation (18). . In the example in Figure 17, the category “Baseball Player” is selected from the dataset with the number of data to be faceted “11401”, and the facet keys “Name”, “Place of Birth”, and “Date of Birth” are selected. Assume that

As illustrated in FIG. 17, on the "new_score" basis, the facet rankings are 31st to 66th, and the MRR is "0.2333". On the other hand, based on the "final_score", the facet rankings are 141st to 211th, and the MRR is "0.0057".

In this way, "name," "place of birth," and "date of birth" are all facets related to "Person" rather than facets specific to "Baseball Player." Therefore, in the "final_score" base in which the degree of association (ont_score) in the ontology is considered, the facet ranking and MRR are lower than in the "new_score" base.

As described above, in the KG 3, the server 2 according to one embodiment performs facet search efficiently by first narrowing down based on the category (class) assigned to the facet, and then narrowing down based on the facet within the category. Realize.

At this time, the server 2 introduces an index that emphasizes class-specific facets, extracts the relationship between facets from the facet search target KG3, and uses this relationship to determine whether the class belongs to a certain class. Efficiently retrieve a set of entities.

Thereby, it is possible to output a plurality of appropriate facets from a set of facets associated with a category, and it is possible to reduce the number of steps it takes for the user to arrive at the desired data.

[1-3] Operation example An operation example of the facet search system 1 described above will be described below with reference to a flowchart.

[1-3-1] DB Creation Process FIG. 18 is a flowchart illustrating an example of the operation of the DB creation process according to one embodiment.

As illustrated in FIG. 18, in the server 2, the frequency table creation unit 23a of the statistical processing unit 23 creates a frequency table 21a (step S1). Furthermore, the score calculation unit 23b of the statistical processing unit 23 creates a score DB 21b based on the frequency table 21a (step S2), and the process ends.

FIG. 19 is a flowchart illustrating an example of the operation of the frequency table creation process in step S1 of FIG. As illustrated in FIG. 19, the frequency table creation unit 23a acquires all classes _Call in KG3 as a frequency table creation process (step S11).

The frequency table creation unit 23a determines whether all classes C, which are elements of _all classes Call, have been processed (step S12). If all have been processed (YES in step S12), the process ends.

If not all have been processed (NO in step S12), the frequency table creation unit 23a calculates the number of instances for each predicate that the instance of class C has, in other words, the facet key frequency (step S13). ), stored in the P frequency table 21a2.

Then, the frequency table creation unit 23a calculates the number of objects (Object), in other words, the facet key/facet value frequency for each predicate held by the instance whose class is C (step S14), and calculates the frequency of the PO frequency table 21a3. , and the process moves to step S12.

FIG. 20 is a flowchart illustrating an example of the operation of the score DB creation process in step S2 of FIG. As illustrated in FIG. 20, as a score DB creation process, the score calculation unit 23b calculates facet frequency, facet balance, key density, and key monotony, respectively, based on the frequency table 21a and KG3 (step S21 ~S24).

The score calculation unit 23b calculates the category importance based on the class and facet information in KG3 (step S25).

Then, the score calculation unit 23b calculates the score org_score shown in the above equation (9) using the facet frequency, facet balance, key density, and key monotonicity (step S26).

Further, the score calculation unit 23b uses the score org_score calculated in step S26 and the category importance calculated in step S25 to calculate a facet score new_score, and stores it in the score DB 21b (step S27), and the process is completed. finish.

[1-3-2] Facet Search Processing FIG. 21 is a flowchart illustrating an operational example of facet search processing according to an embodiment.

As illustrated in FIG. 21, in the server 2, the search control unit 22 receives the selection of class C by the terminal 4 on the category search screen 100 presented to the terminal 4 (step S31).

The semantic interpretation processing unit 24 acquires a predicate held by an instance belonging to class C (step S32).

The semantic interpretation processing unit 24 determines whether all facet keys f, which are elements of the predicate, have been processed (step S33). If not all have been processed (NO in step S33), the semantic interpretation processing unit 24 acquires the score new_score of the facet key f in class C from the score DB 21b (step S34).

The semantic interpretation processing unit 24 uses the class C and the facet key f to calculate a distance score ont_score based on the ontology in the KG3 (step S35), and the process moves to step S33.

In step S33, if all have been processed (YES in step S33), the semantic interpretation processing unit 24 calculates the score final_score of the facet key f using new_score and ont_score (step S36).

The ranking adjustment unit 25 notifies the search control unit 22 of the facet keys sorted based on final_score. The search control unit 22 displays the sorted facet keys in the item list area 210 of the facet search screen 200, which is the transition destination from the category search screen 100 (step S37), and performs the facet search related to the display of the item list area 210. Processing ends.

[2] Others The technique according to the embodiment described above can be modified and changed as follows.

For example, the search control unit 22, statistical processing unit 23 (frequency table creation unit 23a and score calculation unit 23b), meaning interpretation processing unit 24, and ranking adjustment unit 25 included in the server 2 shown in FIG. 4 may be combined in any combination. It may be divided into two parts.

Further, the server 2 shown in FIG. 1 may have a configuration in which a plurality of devices cooperate with each other via a network to realize each processing function. As an example, the search control unit 22 may be a Web server, the statistical processing unit 23, the semantic interpretation processing unit 24, and the ranking adjustment unit 25 may be an application server, the memory unit 21 may be a DB server, etc. In this case, the web server, application server, and DB server may realize each processing function as the server 2 by cooperating with each other via a network.

Furthermore, in one embodiment, the final score (final_score) input to the ranking adjustment unit 25 is obtained by combining the new_score in which signif is taken into account and the ont_score, as shown in the above formulas (8) and (18). The explanation was given as a multiplied score. The final score (final_score) is not limited to this, and at least one of signif and ont_score may be taken into consideration.

For example, the final score (final_score) may match new_score without considering ont_score, as shown in equation (19) below.

final_score(f, C) = new_score(f, C) (19)

Alternatively, the final score (final_score) may be a score that does not take signif into account, as shown in equation (20) below. Note that in one embodiment, the weight ω in the following equation (20) is illustratively “0.5”.

final_score(f, C) = ω * org_score(f, C) + (1 - ω) * ont_score(f, C) (20)

The final score (final_score) shown in the above equation (19) or equation (20) can also produce the same effect as the embodiment illustrated in FIG. 16 or 17.

[3] Additional notes Regarding the above embodiments, the following additional notes are further disclosed.

(Additional note 1)
Accepts specifications for specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; and outputting the plurality of facets in order according to the priority of the plurality of facets calculated based on at least one of the second index based on the distance from
A search program that causes a computer to perform processing.

(Additional note 2)
The first index includes the result of dividing the number of all categories in the knowledge graph by the number of categories with which each of the plurality of facets is associated in the knowledge graph.
Search program described in Appendix 1.

(Additional note 3)
calculating the first index including the results for each of the plurality of facets;
storing information including the calculated first index in a storage area;
The search program according to supplementary note 2, which causes the computer to execute the process.

(Additional note 4)
The second index is based on the distance between the specific category hierarchy in the knowledge graph and each category hierarchy with which each of the plurality of facets is associated in the knowledge graph.
The search program described in any one of Supplementary Notes 1 to 3.

(Appendix 5)
The second index is based on whether each of the plurality of facets is a standard vocabulary.
The search program described in any one of Supplementary Notes 1 to 4.

(Appendix 6)
The outputting process is a search screen that is a transition destination from the screen that accepts the specification of the specific category, and is used to perform a search targeting the plurality of facets associated with the specific category. comprising a process of displaying the plurality of facets arranged in order according to the priority of the plurality of facets on a search screen;
The search program described in any one of Supplementary Notes 1 to 5.

(Appendix 7)
Accepts specifications for specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; and outputting the plurality of facets in order according to the priority of the plurality of facets calculated based on at least one of the second index based on the distance from
A search method in which processing is performed by a computer.

(Appendix 8)
The first index includes the result of dividing the number of all categories in the knowledge graph by the number of categories with which each of the plurality of facets is associated in the knowledge graph.
Search method described in Appendix 7.

(Appendix 9)
calculating the first index including the results for each of the plurality of facets;
storing information including the calculated first index in a storage area;
The search method according to appendix 8, wherein the processing is executed by the computer.

(Appendix 10)
The second index is based on the distance between the specific category hierarchy in the knowledge graph and each category hierarchy with which each of the plurality of facets is associated in the knowledge graph.
The search method described in any one of Supplementary notes 7 to 9.

(Appendix 11)
The second index is based on whether each of the plurality of facets is a standard vocabulary.
The search method described in any one of Supplementary notes 7 to 10.

(Appendix 12)
The outputting process is a search screen that is a transition destination from the screen that accepts the specification of the specific category, and is used to perform a search targeting the plurality of facets associated with the specific category. comprising a process of displaying the plurality of facets arranged in order according to the priority of the plurality of facets on a search screen;
The search method described in any one of Supplementary notes 7 to 11.

(Appendix 13)
A reception department that accepts specifications of specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; an output unit that sequentially arranges and outputs the plurality of facets according to a priority of the plurality of facets calculated based on at least one of a second index based on the distance from the second index;
A search device comprising:

(Appendix 14)
The first index includes the result of dividing the number of all categories in the knowledge graph by the number of categories with which each of the plurality of facets is associated in the knowledge graph.
The search device according to appendix 13.

(Appendix 15)
calculating the first index including the results for each of the plurality of facets;
storing information including the calculated first index in a storage area;
The search device according to supplementary note 14, comprising a calculation unit.

(Appendix 16)
The second index is based on the distance between the specific category hierarchy in the knowledge graph and each category hierarchy with which each of the plurality of facets is associated in the knowledge graph.
The search device according to any one of Supplementary notes 13 to 15.

(Appendix 17)
The second index is based on whether each of the plurality of facets is a standard vocabulary.
The search device according to any one of Supplementary notes 13 to 16.

(Appendix 18)
The output unit is a search screen that is a transition destination from the screen that accepts the specification of the specific category, and the search screen is for performing a search targeting the plurality of facets associated with the specific category. displaying the plurality of facets arranged in order according to the priority of the plurality of facets on a screen;
The search device according to any one of Supplementary notes 13 to 17.

1 facet search system 10 computer 2 server 21 memory section 21a frequency table 21b score DB
22 Search control unit 23 Statistical processing unit 23a Frequency table creation unit 23b Score calculation unit 24 Semantic interpretation processing unit 25 Ranking adjustment unit 3 KG (Knowledge graph)
4 Terminal

Claims (8)

Accepts specifications for specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; and outputting the plurality of facets in order according to the priority of the plurality of facets calculated based on at least one of the second index based on the distance from
A search program that causes a computer to perform processing. The first index includes the result of dividing the number of all categories in the knowledge graph by the number of categories with which each of the plurality of facets is associated in the knowledge graph.
The search program according to claim 1. calculating the first index including the results for each of the plurality of facets;
storing information including the calculated first index in a storage area;
The search program according to claim 2, which causes the computer to execute the process. The second index is based on the distance between the specific category hierarchy in the knowledge graph and each category hierarchy with which each of the plurality of facets is associated in the knowledge graph.
The search program according to any one of claims 1 to 3. The second index is based on whether each of the plurality of facets is a standard vocabulary.
The search program according to any one of claims 1 to 4. The outputting process is a search screen that is a transition destination from the screen that accepts the specification of the specific category, and is used to perform a search targeting the plurality of facets associated with the specific category. comprising a process of displaying the plurality of facets arranged in order according to the priority of the plurality of facets on a search screen;
The search program according to any one of claims 1 to 5. Accepts specifications for specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; and outputting the plurality of facets in order according to the priority of the plurality of facets calculated based on at least one of the second index based on the distance from
A search method in which processing is performed by a computer. A reception department that accepts specifications of specific categories,
a first index based on the number of categories with which each of the plurality of facets associated with the specific category in the knowledge graph is associated, and each of the specific category and the plurality of facets in the knowledge graph; an output unit that sequentially arranges and outputs the plurality of facets according to a priority of the plurality of facets calculated based on at least one of a second index based on the distance from the second index;
A search device comprising:

Images