JP4709671B2 - Knowledge metadata generation apparatus and knowledge metadata generation program - Google Patents

Description

  The present invention relates to a knowledge metadata generation apparatus and a knowledge metadata generation program, and more particularly to a knowledge metadata generation apparatus and a knowledge metadata generation program for generating highly accurate knowledge metadata.

  In recent years, the importance of technology for extracting target information from a vast amount of information has increased. In particular, with respect to a function in which a user directly asks a question to the system and the system generates an answer to the question (so-called Q & A function), it provides useful information even to a user who is not used to handling information equipment. Expectation is gathered from the point that can be.

  Here, in order to realize the above-mentioned Q & A function, a method of searching for an answer to a user's question from a large-scale text corpus such as a WWW (World Wide Web) or a newspaper article is common. For example, by using a huge amount of information resources such as text data on the Internet and article data provided by a newspaper company, it is possible to provide a Q & A function that does not limit the fields in which questions can be answered.

  However, in the Q & A function using a large-scale text corpus such as WWW, for example, there is a possibility that a correct answer cannot be obtained because the reliability of information is not guaranteed. Therefore, conventionally, there is a system that presents an answer and presents it in consideration of information contained in the ground document (extraction source document from which the answer to the question is extracted) (see, for example, Patent Document 1).

  In addition, it is not desirable to use the above-described answer method for a content provider such as a broadcasting station whose mission is to convey correct information. Therefore, a method is conceivable in which highly reliable information (providers can guarantee the contents) is stored in a database in advance and used for answering questions. Note that this method is effective in accurately transmitting reliable information, although the fields that can be answered are limited.

As described above, when extracting information from the database, the user needs to know the structure of the database in advance. However, it is generally impossible for the user to grasp in advance the structure and the like of all the different types of databases that store various information. Therefore, a method for inquiring information without deeply understanding the structure of the database has been proposed (see, for example, Patent Document 2).
Japanese Patent Laid-Open No. 2005-25418 JP 2005-165958 A

  However, in Patent Document 1, which is the conventional method described above, the user who has obtained an answer must make a self-judgment on the reliability of the information group included in the basis document.

  Further, in Patent Literature 2, a query sentence described in an ontology description language is generated for an input natural language sentence using a semantic web technique, and a search using the meaning of information is enabled. However, the actual information is a simple frame representation and does not have a complicated structure, so that highly accurate knowledge data cannot be acquired. Further, Patent Document 2 does not mention how to generate metadata.

  The present invention has been made in view of the above-described problems, and an object thereof is to provide a knowledge metadata generation device and a knowledge metadata generation program for generating highly accurate knowledge metadata.

  In order to solve the above problems, the present invention employs means for solving the problems having the following characteristics.

The invention described in claim 1 is a knowledge metadata generating device that generates knowledge metadata related to content to be provided to a user, and is related to the information resource related to the content and the information resource stored in advance. based on the ontology of the field, and knowledge metadata generation means for generating the knowledge metadata, have a means for storing the knowledge metadata obtained by the knowledge metadata generating means, the knowledge metadata generating means Is obtained from the numerical data and / or structured data obtained as the information resource using the upper-lower structure, partial-whole structure, and cause-result structure of the concept related to the field defined by the ontology. Among the inclusion relationships, upper and lower relationships, and causal relationships between events that occur within Both one is characterized by generating additional knowledge metadata.

According to the first aspect of the present invention, highly accurate knowledge metadata can be generated. That is, information can be knowledge-structured in advance using ontology in a field related to content, and stored as knowledge metadata in a database or the like. Furthermore, more detailed knowledge metadata can be generated by using the relationship between events that occur in the content.

  The invention described in claim 2 is an information extraction unit that extracts inquiry information from input information from the user, and a search unit that searches the knowledge metadata based on inquiry information obtained by the information extraction unit. It is characterized by having.

  According to the second aspect of the present invention, it is possible to search for concepts and the like that are not directly included in the user's request by searching knowledge metadata generated by assuming a question from the user using an ontology. Become. In addition, by using an ontology that defines various conceptual relationships, it is possible to extract flexible answers that match human senses.

According to a third aspect of the present invention, the knowledge metadata generating means relates to the content from structured text data tagged with a subject predicate relationship or dependency related to the content obtained as the information resource. And generating knowledge metadata.

According to the invention described in claim 3 , more detailed knowledge metadata can be generated by using the structured text data tagged with the subject predicate relationship or the dependency relationship related to the content.

According to a fourth aspect of the present invention, there is provided a knowledge metadata generation program for generating knowledge metadata related to content to be provided to a user , wherein the computer includes an information resource related to the content and the information resource stored in advance. Based on the ontology of the field related to the knowledge metadata generating means for generating the knowledge metadata and the storage means for storing the knowledge metadata obtained by the knowledge metadata generating means , The data generation means uses, from the numerical data and / or structured data obtained as the information resource, the upper-lower structure, partial-whole structure, and cause-result structure of the concept related to the field defined by the ontology. , Inclusion relationship between events occurring in the content, upper and lower Relationship, and of the causal relationship, and generating the knowledge metadata at least one which is added.

According to the fourth aspect of the present invention, highly accurate knowledge metadata can be generated. That is, information can be knowledge-structured in advance using ontology in a field related to content, and stored as knowledge metadata in a database or the like. Furthermore, more detailed knowledge metadata can be generated by using the relationship between events that occur in the content. Moreover, knowledge metadata generation can be easily realized by installing an execution program in a computer.

The invention described in claim 5, an information extracting means for extracting the query information from the input information from the previous SL user, based on the inquiry information obtained by said information extraction means, search means for searching said knowledge metadata It is characterized by having.

According to the fifth aspect of the present invention, it is possible to search for concepts and the like that are not directly included in the user's request by searching knowledge metadata generated by assuming a question from the user using an ontology. Become. In addition, by using an ontology that defines various conceptual relationships, it is possible to extract flexible answers that match human senses.

  According to the present invention, highly accurate knowledge metadata can be generated.

<Features of the present invention>
The present invention expresses various pieces of information as structured knowledge metadata by using relationships (ontologies) between concepts that match human senses. In this knowledge metadata, various conceptual relationships assuming questions (inquiries) from users (viewers, etc.) are added to the original information, and by using these relationships, the structure of the database, etc. Inquiry processing that can extract information flexibly without knowing it is possible.

  In addition, knowledge metadata generates structured knowledge data from various highly reliable information resources using an ontology (a collection of clear definitions of relationships between concepts) in a related field of the information resources. Note that the knowledge metadata includes a conceptual relationship assuming a user's question using an ontology.

<Embodiment>
Next, a preferred embodiment of the knowledge metadata generation apparatus and knowledge metadata generation program according to the present invention having the above-described features will be described with reference to the drawings.

  FIG. 1 is a diagram showing an example of a schematic configuration of a knowledge metadata generation system according to the present invention. The knowledge metadata generation system 10 shown in FIG. 1 is configured to include a knowledge metadata generation device 11 and information resource providing devices 12-1 to 12-n. The knowledge metadata generation device 11 and the information resource providing device 12 are connected in a state where data can be transmitted and received via a communication network 13 such as the Internet. The communication network 13 may be wired or wireless.

  The knowledge metadata generation device 11 uses a predetermined ontology from one or a plurality of ontologies stored in the device or outside in advance to obtain knowledge from information resources such as numerical data, structured data, and text data related to the content. Generate metadata. The knowledge metadata generation apparatus 11 acquires the above-described information resources from the information resource providing apparatuses 12-1 to 12-n connected to the communication network 13 or the like. Further, when generating knowledge metadata, the knowledge metadata generation device 11 relates to the content contents from, for example, structured text data related to a predetermined content tagged with a subject predicate relationship, dependency relationship, etc. The generated knowledge metadata is generated automatically or semi-automatically (for example, partly including manual work).

  Furthermore, the knowledge metadata generation device 11 may have a display output function for providing content to the user, in which case the user has a request for any question regarding the content being viewed or watched. In this case, an appropriate answer can be extracted from a plurality of generated knowledge metadata and provided to the user. The request from the user includes, for example, inquiry information described in a predetermined format or the like, or inquiry information extracted from a user's question sentence given by a natural language sentence.

  The information resource providing devices 12-1 to 12-n accumulate information resources related to content. Further, the information resource providing devices 12-1 to 12-n provide the information resources to the knowledge metadata generating device 11 via the communication network 13. The information resource providing method in the information resource providing apparatus 12 may be provided when the information resource providing apparatus 12 receives an acquisition request from the knowledge metadata generating apparatus 11, for example. It may be provided at a predetermined timing such as an update of information resources.

  The information resource stored in the information resource providing device 12 is a highly reliable information resource provided by, for example, a service provider, a data distribution company, a newspaper company, a broadcasting station, etc., and each information resource is an information resource providing device. It is accumulated in any of 12-1 to 12-n. The resource information provider provided in the present invention is not limited to this.

<Knowledge metadata generation device 11>
Next, the knowledge metadata generation device 11 will be specifically described with reference to the drawings. FIG. 2 is a diagram illustrating an example of a functional configuration of the knowledge metadata generation apparatus. The knowledge metadata generation apparatus 11 shown in FIG. 2 includes an input means 21, an output means 22, a transmission / reception means 23, a storage means 24, a knowledge metadata generation means 25, an information extraction means 26, and a search means 27. , And control means 28.

  The input unit 21 accepts input of content output start / end instructions, inquiries about content, and the like from the user. Note that the input unit 21 includes a pointing device such as a keyboard and a mouse, a voice input interface such as a microphone, and the like.

  The output unit 22 displays and / or outputs the instruction content input by the input unit 21, the content reproduced based on the instruction content, and the answer to the inquiry information by voice. Note that the output means 22 includes a display, a speaker, and the like.

  The transmission / reception means 23 is a communication interface for acquiring content via the communication network 13. Further, the transmission / reception means 23 acquires information resources that are the basis of the knowledge base from the information resource providing devices 12-1 to 12-n described above via the communication network 13. The transmission / reception means 23 can also acquire an ontology from an external device (not shown) connected to the communication network 13 and provide the generated knowledge base to other terminals.

  The storage unit 24 includes an information resource 31 acquired by the transmission / reception unit 23, an ontology 32 related to a preset field, and a knowledge base (knowledge database) 33 which is a knowledge metadata group generated using the ontology 32. Accumulate. The storage unit 24 can also store content to be output to the user.

  Further, various data stored in the storage unit 24 can be read out and written as necessary. Furthermore, the storage means 24 is systematically configured so that information on various stored data can be searched.

  Based on the information resource 31, the knowledge metadata generation unit 25 generates a knowledge base 33 including at least one knowledge metadata using the ontology 32. The knowledge metadata generation unit 25 quickly generates the knowledge base 33 as soon as information resources corresponding to the content provided to the user can be acquired. At this time, the knowledge metadata generation unit 25 transmits an acquisition request for the information resource corresponding to the predetermined content to the information resource providing apparatus 12 connected via the communication network 13 by the transmission / reception unit 23, and the predetermined information resource May be obtained. Thereby, it is possible to quickly respond to an inquiry from the user. An example of ontology generation in the knowledge metadata generation unit 25 will be described later.

  In addition, the information extraction unit 26 extracts a portion corresponding to the inquiry information from input information such as a character string or speech input by the input unit 21. As an information extraction method in the information extraction means 26, for example, a word used for a preset search is extracted from a character string, or inquiry information is extracted by a conventional language processing system such as dependency analysis.

  The search means 27 searches the knowledge base 33 stored in the storage means 24 based on the contents extracted by the information extraction means 26, generates response data from the extracted contents, and outputs the response data via the output means 22. Present it to the user. An information (answer) extraction method from knowledge metadata in the search means 27 will be described later.

  The control means 28 controls the entire functional configuration in the knowledge data processing apparatus 11. Specifically, the control unit 28 causes the information extraction unit 26 to extract inquiry information from the input information triggered by the input information from the user input by the input unit 21, and responds from the knowledge base 33 by the search unit 27. Control is performed such as extracting the answers and outputting the answer data, contents, and the like by the output means 22.

  Here, in the knowledge metadata generation device 11 described above, various data stored in the storage unit 24 is stored in an external device (not shown) connected to the communication network 13 or a storage unit included in the information resource providing device 12. The data may be accumulated and accessed as necessary to obtain predetermined data, or various data may be accumulated.

<About knowledge metadata>
Here, knowledge metadata in the present invention will be described. Knowledge metadata is structured knowledge data generated from various information resources (for example, numerical data, structured data such as XML, normal text data, etc.) using ontology in related fields. By limiting information resources to reliable ones, accurate and reliable knowledge metadata can be generated.

  At the time of generating the knowledge metadata, an ontology vocabulary regarding a certain field prepared in advance is used to add a conceptual relationship assuming a user question. As a result, it is possible to search for information related to concepts that are not directly included in the user's question or to search for information using general common sense regarding the field.

  In order to realize information retrieval in the present invention, for example, an ontology description language OWL (Ontology Web Language), a knowledge base query language RDQL (RDF Data Query Language), and various semantic web related tools are used. Can be realized.

<Example 1>
Next, a specific embodiment of the knowledge metadata generation apparatus 11 will be described. FIG. 3 is a diagram illustrating a schematic example of the knowledge metadata generation apparatus according to the first embodiment. Here, the knowledge metadata generation apparatus 11 according to the present invention can be implemented in an information provider (information resource) server, an Internet public server, a PC terminal in a user's home, or a TV receiver. In the first embodiment shown in FIG. 3, an example of a system configuration when mounted in the TV receiver 41 is shown.

  The knowledge metadata generation device 11 in the TV receiver 41 in the viewer (user) 's home has accumulated knowledge base (kb) for accumulating knowledge metadata (kmd) and knowledge related to program contents. An ontology (ont) is provided. The knowledge metadata generation apparatus 11 receives data d (numerical data, structured data such as XML) and data a (text data) from at least one highly reliable content provider such as an external data distribution company or a newspaper company. Or the like, and convert it into knowledge metadata in accordance with the format of the schema (s) provided by the ontology (ont). The schema (s) is a template that defines how to describe information when structuring information.

  Further, the above-described conversion process may be performed semi-automatically by hand for data of a type that requires analysis processing, but may be performed automatically depending on the type of data. The generated knowledge metadata (kmd) is stored in the knowledge base (kb). The user can extract information from the knowledge base using a dedicated query language or natural language sentence. Hereinafter, an example of ontology, an example of knowledge metadata description, and an example of information extraction from knowledge metadata will be described.

<Example of ontology>
FIG. 4 is a diagram illustrating an example of an ontology that provides a schema for generating knowledge metadata. FIG. 4 shows a part of an ontology describing a concept relating to baseball. That is, in FIG. 4, general knowledge about a baseball game held by humans is structured.

  Here, in order to structure a plurality of concepts related to baseball, events such as “throw”, “hit”, and “strike” are expressed as concepts (classes), and the “higher-lower structure (for example, is_a)” of the class And “partial-whole structure (for example, hasEvent, isEventOf, etc.)”, “cause-result structure (for example, causatedBy, cause, etc.)” and the like are defined as relationships (properties).

  Such a graph structure can be described by an ontology language such as OWL. Further, knowledge metadata representing the actual game situation is described as an instance of these classes.

<Example description of knowledge metadata>
FIG. 5 is a diagram showing an example of knowledge metadata described using the baseball ontology shown in FIG. The example shown in FIG. 5 is knowledge metadata describing the situation of “running a curve ball and 2 runners” in a baseball game. Here, FIG. 5A shows an example of knowledge metadata description by OWL, and FIG. 5B shows its graph structure.

  In FIG. 5, at a certain inning (inning02102) of a game, a certain bat (atBat0210206) occurs, and a throwing event (curve0210206_1), a batting event (double0210206_2), and a running event (run0210206_3) occur at that bat. This is shown using the isEventOf property.

  Furthermore, it is indicated by using the coordinatedBy property that a hitting event (double0210206_2) occurs due to a pitching event (curve0210206_1) and a striking event (run0210206_3) occurs due to the hitting event.

  The knowledge metadata generation means 25 can construct a knowledge base related to a baseball game by accumulating a large number of knowledge metadata describing relationships as shown in FIG.

<Extraction of information (answer) from knowledge metadata in search means 27>
Next, an information (answer) extraction method from knowledge metadata in the search means 27 will be described with reference to the drawings. FIG. 6 is a diagram illustrating an example of information extraction from knowledge metadata. For extracting information from the knowledge metadata, an inquiry language such as RDQL can be used as shown in FIG. RDQL is a query language having a structure similar to SQL, and is a language for searching a database of RDF (Resource Description Framework) at high speed. Note that RDF is a specification for describing meta information on the Semantic Web. In the present invention, it is possible to extract information using inference by using knowledge metadata (instance) and ontology (schema).

  For example, in the example of FIG. 6, RDQL corresponding to an inquiry example “who has hit two times?” Is shown. Here, in an actual game, it is assumed that there is a player who has made two strikes twice. In the knowledge metadata, it is assumed that a player who has made a single hit is not described, and only a player who has made a second hit is described.

  On the ontology, since the second hit is defined as a subordinate concept of the hit, the inference engine for processing this RDQL can extract not only a single hit but also a second hit player as an answer.

  In other words, as a human sense, there is a common sense that the player who hit the hit is the player who hit the hit, and as a response to the question, it is reasonable to search for all types of hits, not just hits. This is based on the idea.

  FIG. 7 is a diagram illustrating an example of information extraction from knowledge metadata in a more complicated inquiry example. FIG. 7 shows an RDQL corresponding to an example of an inquiry from the user “who was hit twice?”. Here, in the example shown in FIG. 7, in response to an inquiry “Who was hit twice?”, For example, an inference using the following relationship is performed.

<< Concept of upper and lower concepts >>
Actually, the batter hits twice, but using the fact that the double hit class is a subordinate concept of the hit class, this double hit can be obtained as a search result. In other words, the user generally uses the expression “hit a hit” to mean “hit a hit”, and naturally, the hit event at that time is a home run or a two-stroke hit. An event such as (Double) is also considered to be included. By considering the upper-lower relationship of such an event, it is possible to search for a two-stroke or home run that is a lower-level concept for a query for “hit”.

<< Use of transitivity of relationships >>
Here, in the baseball ontology described above, the isEventOf relationship is defined as “Transitive”. In other words, it is described that “a certain double strike is an event that occurred at a specific at-bat”, and “a specific batting is an event that occurred at a specific inning” is described, It can be inferred that the event occurred at a specific inning.

  As a result, even if there is no direct description, it can be seen that, for example, two strikes have occurred twice.

<< Cause-Result Relationship >>
For example, “the pitcher who hit the hit” is “the player who made the pitch that caused the hit”. Therefore, by applying RDQL using a causal relationship (causedBy) as shown in FIG. It is possible to search for information of players who have thrown a hit curve.

  Also, if the “causedBy” property is also defined as “transitive”, it is possible to flexibly answer complex inquiries such as “Which pitcher did you score? Is possible.

  As described above, in this embodiment, by using the relationship between concepts described in the baseball ontology, it becomes possible for the user to perform a search by the RDQL query without knowing the detailed structure in the knowledge base. . That is, information extraction using inference can be realized by using an ontology structure such as a concept upper-lower relationship, a whole partial relationship, a cause-result relationship, and the like. Therefore, by performing inference processing using the relationship between concepts, a flexible question answer that cannot be realized by conventional surface word search is possible.

<Difference between conventional content metadata and knowledge metadata>
Here, the difference between the conventional content metadata and the knowledge metadata in the present invention will be described with reference to the drawings. FIG. 8 is a diagram illustrating an example for explaining the difference between conventional content metadata and knowledge metadata. First, the conventional content metadata shown in FIG. 8A is data in which information such as what is reflected in the video or what is happening is described by performing video recognition, voice recognition, event recognition, and the like. is there.

  For example, in the case of sports such as baseball, information such as who is standing at a batting point and what scene is described. Such information can be used for digest generation and event extraction.

  On the other hand, the knowledge metadata in the present invention in FIG. 8B also describes the relationship (inclusion relationship or causal relationship) between the events that have occurred. All of these are added on the assumption that the viewer (user) asks questions about the content.

  As described above, in the present invention, the knowledge metadata generation means 25 generates knowledge metadata to which at least one of the inclusion relationship between the events occurring in the content, the upper and lower relationship, and the causal relationship is added from the information resource. By doing so, it is possible to generate more detailed knowledge metadata using the relationship between events that occur in the content. Therefore, flexible Q & A processing such as searching for a concept that is not directly included in the user's question can be realized.

<Example 2>
Next, another embodiment (embodiment 2) to which the present invention is applied will be described with reference to the drawings. FIG. 9 is a diagram illustrating a schematic example of the knowledge metadata generation apparatus according to the second embodiment. In the second embodiment, an implementation example of a Q & A system for sports programs using the present invention will be described. Further, in the second embodiment, it is assumed that a user who is watching a baseball game on TV temporarily takes a seat by a visitor or a telephone and asks the game situation when returning to the TV. To do.

  Sports data distribution companies and the like provide game data consisting of numerical values and structured data (sd) in a unique format for the progress of the game for each pitch. Currently, broadcasting stations and sports-related websites use these information to provide users with game status such as SBO counts, scores, ball types, and hit results.

  Broadcasting stations that conduct baseball broadcasts provide comments and commentary from announcers and commentators, along with video and audio of the game. These are distributed to the user as text data (td) such as caption broadcasting.

  The knowledge metadata generation device 11 connected to the TV receiver 41 or to the communication network 13 generates knowledge metadata (s-kmd) from structured data (sd) received from a sports data distribution company or the like. At that time, metadata generation processing is performed using the vocabulary of the baseball ontology stored in the knowledge metadata generation device 11 or an external device such as a server. In addition to the progress of the game, static information such as the player's career and results may be converted into knowledge metadata as well, and this knowledge metadata can also be used to generate an answer to the user.

  Here, FIG. 10 is a diagram illustrating an example of knowledge metadata generation from game information. For example, for a specific field such as the baseball field, the inclusion relationship, the causal relationship, the upper and lower relationship, etc. between events occurring during a game are clear. Therefore, as shown in FIG. 10, the baseball ontology including the baseball ontology is set in advance from structured data (sd) including game information for each ball (pitcher: Watanabe, pitch: curve, batter: Imaoka, result: double strike, etc.). It is possible to convert into structured knowledge metadata (s-kmd) using sports ontology (s-ont).

  Also, in FIG. 9, text data (td) such as commentator comments is provided in a language processing system (nlp) in the broadcasting station side, the receiver side, or the knowledge metadata generation device 11 and the language processing system (nlp). Is converted into structured text data (s-td). The structured text data (s-td) is a text document tagged with a subject-predicate relationship, a subject-predicate-object relationship, a dependency relationship, and the like of words. The knowledge metadata generation device 11 can convert the structured text data (s-td) into knowledge metadata (s-kmd) using the tag information of the structured text data (s-td). As a result, more detailed knowledge metadata can be generated by using structured text data in which subject predicate relationships, dependency relationships, etc. are tagged in relation to content.

  The generated knowledge metadata is stored in a knowledge base (s-kb) stored in the knowledge metadata generation device 11 or an external device such as a server connected to the communication network 13. The viewer watching the TV generates the query language (RDQL) as shown in FIG. 6 or FIG. 7 from the utterance through the natural language processing system or the character information input through a specific interface and the like. Information (answer) can be extracted from the base (s-kb).

  FIG. 11 is a diagram showing an example of knowledge metadata generation from text information. As shown in FIG. 11, when “reversed by Kiyohara's home run” is input as text data (td), the structured text data (s-td) is converted by a preset language processing system (nlp) or the like. Knowledge metadata is generated using the structured text data (s-td).

<Example 3>
Next, another embodiment (Example 3) to which the present invention is applied will be described. Embodiment 3 describes an implementation example of a Q & A system for information programs to which the present invention is applied. FIG. 12 is a diagram illustrating a schematic example of the knowledge metadata generation apparatus according to the third embodiment.

  In the third embodiment, it is assumed that a user who auditioned an information program on health on TV wants to acquire information on prevention and treatment of diseases introduced in the program after viewing.

  As shown in FIG. 12, the broadcasting station 42 uses structured text data (td) from the text data (td) uttered by a performer (eg, a doctor or an announcer) in the program by a language processing system or manual work. s-td).

  The knowledge metadata generation device 11 uses a health (medical) ontology (m-ont) 32 stored in the knowledge metadata generation device 11 or an external device such as a server connected to the communication network 13. Then, knowledge metadata (m-kmd) is generated and stored in the knowledge metadata generation device 11 or in the knowledge base (m-kb) 33 of the external device.

  Here, FIG. 13 is a diagram illustrating an example of a health ontology. FIG. 13 shows an example of an ontology related to health used in a health program. FIG. 14 is a diagram illustrating an example of knowledge metadata generation in a health program. That is, FIG. 14 shows an example of knowledge metadata generated from the utterance contents in the program based on the health ontology shown in FIG. Therefore, the knowledge metadata generation device 11 uses the knowledge metadata shown in FIG. 14 and is input by an utterance or a specific interface through the natural language processing system regardless of whether or not the viewer is watching the program. A query language (RDQL) as shown in FIG. 6 or FIG. 7 described above can be generated from character information or the like, and information can be extracted from the knowledge base (m-kb).

  As shown in the above-described embodiments, the present invention can knowledge-structure information in advance using ontology in a field related to content, and store the information in a database or the like as knowledge metadata. As a result, it becomes possible for the user to obtain only reliable information. In addition, by generating knowledge metadata that assumes a user's question, it is possible to search for concepts that are not directly included in the user's request.

<Execution program>
Here, the above-described knowledge metadata generation device can perform knowledge metadata generation (knowledge metadata generation) in the present invention using the above-described dedicated device configuration or the like, but executes the processing in each configuration on a computer The knowledge metadata generation according to the present invention can be realized by generating an execution program for causing the program to be executed and installing the program in, for example, a general-purpose personal computer or server.

<Hardware configuration>
Here, a hardware configuration example of a computer capable of executing knowledge metadata generation according to the present invention will be described with reference to the drawings. FIG. 15 is a diagram illustrating an example of a hardware configuration capable of realizing knowledge metadata generation processing according to the present invention.

  15 includes an input device 51, an output device 52, a drive device 53, an auxiliary storage device 54, a memory device 55, a CPU (Central Processing Unit) 56 that performs various controls, and a network connection device. 57 are connected to each other by a system bus B.

  The input device 51 includes a keyboard and a pointing device such as a mouse operated by the user, a voice input device, and the like, and inputs various operation signals and voice signals such as a program execution instruction from the user. The output device 52 has a display, a speaker, and the like that display various windows and data necessary for operating the computer main body for performing the processing in the present invention. Display or audio output is possible.

  Here, in the present invention, the execution program installed in the computer main body is provided by a recording medium 58 such as a CD-ROM. The recording medium 58 on which the program is recorded can be set in the drive device 53, and the execution program included in the recording medium 58 is installed in the auxiliary storage device 54 from the recording medium 58 via the drive device 53.

  Further, the drive device 53 can record the execution program according to the present invention in the recording medium 58. Thereby, using the recording medium 58, it can be easily installed in a plurality of other computers, and knowledge metadata generation processing can be easily realized.

  The auxiliary storage device 54 is a storage means such as a hard disk, and can store an execution program according to the present invention, a control program provided in a computer, etc., and perform input / output as necessary. The auxiliary storage device 54 can also be used as a storage unit that stores the ontology, knowledge base, various information resources, and the like described above.

  Based on a control program such as an OS (Operating System) and an execution program read from the auxiliary storage device 54 and stored in the memory device 55, the CPU 56 inputs various calculations and data with each hardware component. Each processing in the knowledge metadata generation processing can be realized by controlling processing of the entire computer such as output. Various information necessary during the execution of the program can be acquired from the auxiliary storage device 54 and stored.

  The network connection device 57 obtains an execution program from another terminal connected to the communication network or executes the program by connecting to a communication network such as a telephone line or a LAN (Local Area Network) cable. The execution result obtained in this way or the execution program in the present invention can be provided to other terminals or the like.

  With the hardware configuration as described above, knowledge metadata generation processing can be realized at low cost without requiring a special device configuration. In addition, knowledge metadata generation processing can be easily realized by installing a program.

<Knowledge metadata generation processing procedure>
Next, a knowledge metadata generation processing procedure using an execution program (knowledge metadata generation program) in the present invention will be described with reference to a flowchart. FIG. 16 is a flowchart illustrating an example of a knowledge metadata generation processing procedure.

  In FIG. 16, first, information resources such as numerical data, structured data, and text data related to content are input (S01). Next, knowledge metadata is generated based on the ontology of the field related to the information resource stored in advance for the input information resource (S02). Further, knowledge metadata generated in the process of S02 is accumulated (S03). Note that the above-described processes of S01 to S03 are executed in advance before the processes of S04 to S07 described later, and are accumulated in an accumulation unit or the like.

  Next, input information from a user (viewer) who is viewing the content is acquired (S04), and inquiry information is extracted from the information (S05). In this case, the inquiry information may be extracted from the input information by a preset natural language processing system, or may be specifically input to the viewer as inquiry information through an interface or the like.

  Next, based on the extracted inquiry information, knowledge metadata is searched, an answer corresponding to the inquiry is extracted (S06), and the extracted answer data is presented to the viewer (S07).

  As described above, highly accurate knowledge metadata can be generated by knowledge metadata generation processing using a knowledge metadata generation program. In addition, knowledge metadata generation processing can be easily realized by installing a program.

  As described above, according to the present invention, highly accurate knowledge metadata can be generated. Specifically, according to the present invention, information can be knowledge-structured in advance using ontology in a field related to content, and stored as knowledge metadata in a database or the like. As a result, it becomes possible for the user to obtain only reliable information. In addition, by generating knowledge metadata that assumes a user's question, it is possible to search for concepts that are not directly included in the user's request. In addition, the present invention can extract information (answers) flexibly according to human senses by using an ontology that defines various conceptual relationships.

The present invention can also be applied to event extraction and digest generation in content. For example, in a baseball ontology, a class with a specific property and property value is a “definition class”.
As described above, a specific instance can be extracted from knowledge metadata. Therefore, by defining on the ontology the event such as “reversal” in the game as “scoring event in which the point difference situation reverses before and after the occurrence”, it is more complicated event from the game situation. Can be extracted.

  In addition, by analyzing the program creator's digest production method, describing the features of important scenes with ontologies, and defining concepts such as the appearance of a smasher with a full count and the chance of a goodbye, It can be applied to digest scene extraction and the like. In addition, the content in this invention is not limited to a program, It can apply to all content, such as a moving image, a music, a movie, and text information.

  Although the preferred embodiment of the present invention has been described in detail above, the present invention is not limited to the specific embodiment, and various modifications are possible within the scope of the gist of the present invention described in the claims. Can be changed.

It is a figure which shows an example of schematic structure of the knowledge metadata production | generation system in this invention. It is a figure which shows an example of a function structure of a knowledge metadata production | generation apparatus. It is a figure which shows the example of an outline | summary of the knowledge metadata production | generation apparatus in Example 1. FIG. It is a figure which shows an example of ontology which provides the schema at the time of producing | generating knowledge metadata. It is a figure which shows an example of the knowledge metadata described using the baseball ontology shown in FIG. It is a figure which shows an example of the information extraction from knowledge metadata. It is a figure which shows the example of information extraction from the knowledge metadata in the more complicated inquiry example. It is a figure of an example for demonstrating the difference between the conventional content metadata and knowledge metadata. It is a figure which shows the example of an outline | summary of the knowledge metadata production | generation apparatus in Example 2. FIG. It is a figure which shows an example of the knowledge metadata production | generation from game information. It is a figure which shows the example of knowledge metadata production | generation from text information. It is a figure which shows the example of an outline | summary of the knowledge metadata production | generation apparatus in Example 3. FIG. It is a figure which shows an example of health ontology. It is a figure which shows the example of knowledge metadata production | generation in a health program. It is a figure which shows an example of the hardware constitutions which can implement | achieve the knowledge metadata production | generation process in this invention. It is a flowchart which shows an example of a knowledge metadata production | generation processing procedure.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Knowledge metadata generation system 11 Knowledge metadata generation apparatus 12 Information resource provision apparatus 13 Communication network 21 Input means 22 Output means 23 Transmission / reception means 24 Storage means 25 Knowledge metadata generation means 26 Information extraction means 27 Search means 28 Control means 31 Information Resource 32 Ontology 33 Knowledge Base 41 TV Receiver 42 Broadcast Station 51 Input Device 52 Output Device 53 Drive Device 54 Auxiliary Storage Device 55 Memory Device 56 CPU
57 Network connection device 58 Recording medium

Claims (5)

In a knowledge metadata generation device that generates knowledge metadata related to content provided to a user,
Knowledge metadata generation means for generating the knowledge metadata based on an information resource related to the content and an ontology of a field related to the information resource accumulated in advance;
Have a means for storing the knowledge metadata obtained by the knowledge metadata generation means,
The knowledge metadata generation means includes
From the numerical data and / or structured data obtained as the information resource, using the upper-lower structure, partial-whole structure, and cause-result structure of the concept related to the field defined by the ontology, in the content A knowledge metadata generation device that generates knowledge metadata to which at least one of an inclusion relationship, an upper-lower relationship, and a causal relationship between generated events is added . Information extracting means for extracting inquiry information from the input information from the user;
The knowledge metadata generation apparatus according to claim 1, further comprising a search unit that searches the knowledge metadata based on inquiry information obtained by the information extraction unit. The knowledge metadata generation means includes
From subject predicate relations or dependencies tagged structured text data associated with the content obtained as the information resource to claim 1 or 2, characterized in that to generate the knowledge metadata associated with the content Description knowledge metadata generation device. In a knowledge metadata generation program that generates knowledge metadata related to content provided to a user,
Computer
Knowledge metadata generation means for generating the knowledge metadata based on information resources related to the content and ontology in a field related to the information resources accumulated in advance ; and
Function as storage means for storing knowledge metadata obtained by the knowledge metadata generation means ;
The knowledge metadata generation means includes
From the numerical data and / or structured data obtained as the information resource, using the upper-lower structure, partial-whole structure, and cause-result structure of the concept related to the field defined by the ontology, in the content A knowledge metadata generation program that generates knowledge metadata to which at least one of an inclusion relationship, an upper-lower relationship, and a causal relationship between generated events is added . And information extraction means for extracting the query information from the input information from the previous Symbol user,
5. The knowledge metadata generation program according to claim 4 , further comprising search means for searching the knowledge metadata based on inquiry information obtained by the information extraction means .

Images