JP6499477B2 - Ontology generation device, metadata output device, content acquisition device, ontology generation method, and ontology generation program - Google Patents

Description

  The present invention relates to an ontology generation apparatus, method, and program.

With the spread of video distribution services, the importance of video content has increased, and the use of video content has advanced in various fields, and the amount of content has also increased.
For this reason, it is desirable that metadata that clarifies what the content is is given to the content. The content holder can present the content of the video content to the user in text by adding metadata such as “This content is a spawning image of the organism A”. Also, for example, by defining “organism A” and “laying eggs” separately, the content holder can present not only other contents related to “living organism A” but also a plurality of living images related to “laying eggs” in association with each other.
Patent Document 1 proposes a method of searching for a specific community site using a title assigned to video content as a search keyword, extracting information from the obtained Web page, and assigning it to video content as metadata. .

JP 2008-4080 A

By the way, it is necessary for a person who inputs metadata to input metadata by imagining a scene or service to be used. For example, only “living organism A” may be input, or “spawning of living organism A” may be input. Moreover, the definition of metadata differs depending on the input person even if the content is the same. For example, metadata “spawning” of “organism A” may be input, and “reproduction” of “organism A” may be input.
In the method of Patent Document 1, one web page of a community site describing video content is found, and information is extracted from this web page. Since the structure of the metadata to be assigned depends on a single Web page, the definition of the metadata to be assigned may be different for different video contents.

  It is an object of the present invention to provide an ontology generation apparatus, method, and program capable of generating knowledge structure data that defines a metadata system that can be commonly used by a plurality of contents.

  An ontology generation device according to the present invention includes a headline extraction unit that extracts hierarchical information of headwords from a plurality of document information in a specified field, and a word extraction unit that extracts words associated with the headwords from the document information And a structuring unit that integrates the headwords based on the similarity of the words and generates knowledge structure data of the field including hierarchical information of the integrated headwords.

  The structuring unit may generate knowledge structure data in which the headword and the word are associated with each other.

  The structuring unit may generate the knowledge structure data including a degree of association between the headword and the word.

  The metadata output device according to the present invention provides hierarchical information of the headword associated with the word by matching the text data related to the content with the word included in the knowledge structure data generated by the ontology generation device. An output unit is provided that extracts the knowledge structure data and outputs it as metadata of the content.

  The metadata output device includes a dictionary acquisition unit that acquires a synonym of the word based on dictionary data, and the output unit applies the word to the word by matching the text data with the word or the synonym. You may extract the hierarchical information of the said headword linked | related.

  The content acquisition device according to the present invention includes a first content acquisition unit that acquires, from a predetermined database, content to which the same metadata as the metadata output by the metadata output device is assigned.

  The content acquisition apparatus extracts, from the knowledge structure data, another hierarchical information having a common upper level of the hierarchical information based on the hierarchical information of the metadata, and is given metadata corresponding to the different hierarchical information. A second content acquisition unit that acquires the acquired content from the predetermined database may be provided.

  In the ontology generation method according to the present invention, a computer control unit extracts a headword hierarchical information from a plurality of document information in a specified field, and a word associated with the headword is extracted from the document information. A word extracting step for extracting from the word, and a structuring step for integrating the headwords based on the similarity of the words and generating knowledge structure data of the field including hierarchical information of the integrated headwords. Run.

  The ontology generation program according to the present invention includes a headline extraction step of extracting hierarchical information of headwords from a plurality of pieces of document information in a specified field, and a word associated with the headword in the document information. A word extracting step for extracting from the word, and a structuring step for integrating the headwords based on the similarity of the words and generating knowledge structure data of the field including hierarchical information of the integrated headwords. Let it run.

  According to the present invention, it is possible to generate knowledge structure data that defines a metadata system that can be commonly used by a plurality of contents.

It is a figure which shows the function structure of the management server which concerns on 1st Embodiment. It is a figure which shows an example of the production | generation process of ontology which concerns on 1st Embodiment. It is a figure which shows an example of ontology which concerns on 1st Embodiment. It is a figure which shows an example of the output method of the metadata which concerns on 1st Embodiment. It is a flowchart which shows an example of the provision process of the metadata which concerns on 1st Embodiment. It is a figure which shows the function structure of the content server which concerns on 2nd Embodiment. It is a figure which shows an example of the acquisition process of the related content which concerns on 2nd Embodiment.

[First Embodiment]
The first embodiment of the present invention will be described below.
The management server 1 according to the present embodiment is an information processing apparatus (computer) that is accessed directly or via a network by a content holder, a service provider, or the like. The management server 1 functions as an ontology generation device that builds ontology as knowledge structure data from information on a community site, and a metadata output device that assigns metadata based on the ontology to content such as video or Web pages.

FIG. 1 is a diagram illustrating a functional configuration of the management server 1 according to the present embodiment.
The management server 1 includes a target designation unit 11, a headline extraction unit 12, a word extraction unit 13, a structuring unit 14, a dictionary acquisition unit 15, a content designation unit 16, and an output unit 17.

  The target specifying unit 11 acquires document information that is a target for extracting a knowledge structure in a certain field (for example, “living organism”). The document information is, for example, a Web page that explains individual headwords on a community site such as Wikipedia (registered trademark). The target designating unit 11 receives the designation input of the URL of the Web page and acquires page data, or directly accepts document information from a user as a file such as CSV. By specifying a plurality of extraction targets belonging to “living organisms” as extraction targets, the target designating unit 11 can acquire document information for obtaining characteristics of the “living organisms” as a whole. Note that even if a single Web page can be used as it is as a knowledge structure, the extraction target may be single.

  The headline extraction unit 12 analyzes the structure of headwords in each of a plurality of pieces of document information acquired by the target specification unit 11, and extracts headwords and hierarchical information thereof from these document information.

The word extraction unit 13 extracts a word associated with the headword from the document information.
For example, the word extraction unit 13 performs morphological analysis on the document information, and uses a word having high importance based on a predetermined index (for example, TF-IDF) among the obtained words in each heading. It may be extracted as a characteristic word.

The structuring unit 14 integrates headwords based on the word similarity, and generates an ontology including hierarchical information of the integrated headwords for each field.
For example, in the field of living organisms, each document information is explained by unconventional headwords such as “feature”, “habitat”, “distribution”, “life history”, “reproduction”, etc. The unit 14 generates a common and systemized ontology by integrating headwords having a high degree of similarity.

  In the ontology, headwords and words are associated with each other. Furthermore, the ontology includes the degree of association between a headword and a word based on the appearance probability of the same word for a plurality of headwords.

FIG. 2 is a diagram illustrating an example of an ontology generation process according to the present embodiment.
First, the target designating unit 11 acquires text data of a web page such as “daphnia”, “sea urchin”, “dragonfly”, etc. as document information related to a living organism according to the designation of the URL.

  From the acquired text data, the headline extraction unit 12 uses, for example, “internal” and “external” below “morphology” and “food”, “reproductive” and “low” below “ecology” as hierarchical information of headwords. Extract structure data associated with each "distribution".

  The word extraction unit 13 extracts a word characterizing this headword from the range of text data corresponding to each headword. For example, words such as “eat” and “food” are extracted for the headword “food”.

  The structuring unit 14 integrates different headwords into one from the similarity of the words in these headlines, and generates hierarchical information of the headwords that are made common. For example, when two headwords of “internal form” and “internal structure” are extracted, since the words associated with them are matched or similar with high probability, the two headwords are integrated into one.

The dictionary acquisition unit 15 acquires synonyms of words included in the ontology based on the dictionary data.
For example, the dictionary acquisition unit 15 acquires a similar term having a similar conceptual structure by adding a conceptual structure to an ontology word using dictionary data.

FIG. 3 is a diagram illustrating an example of an ontology to which a conceptual structure according to the present embodiment is added.
This ontology holds the structure and necessary words necessary to explain “living organisms”. For example, the concept of “external” can define a conceptual structure having “form” as a superordinate concept and having a verb “shrink” and a noun “arm” as instances. In addition, from these sets and dictionary data, the “form-external” concept is defined as having a noun related to “body”, and therefore, a relationship such as “nose” or “foot” having the same conceptual structure as “arm”. Even if a noun appears, it can be classified as “form-external”.
The ontology is described using a description language such as OWL.

The content specifying unit 16 acquires text data regarding the content to which metadata is to be added.
The content is, for example, video content of a content holder, and the content specifying unit 16 extracts text information of the specified content or accepts input of text itself to which metadata is to be added.

The output unit 17 decomposes text data related to the content into morphemes, and then matches these morphemes with words or similar terms included in the ontology. Based on the result of this matching, the hierarchical information of the headwords associated with the words having high similarity to the text data of the content is extracted from the ontology and output as metadata representing the content content.
At this time, the output unit 17 extracts the hierarchical information of the headword having a higher score calculated based on the degree of association between the matched word and the headword.

FIG. 4 is a diagram illustrating an example of a metadata output method based on the score according to the present embodiment.
In this example, words such as “back”, “nurturing”, “egg”, and “laying” are extracted from the text data related to the content. By matching these with knowledge structure data, a score for each metadata is calculated.

  For example, a score corresponding to the degree of association is added to “inside”, “outside”, “food habit”, and “reproductive” of the headwords matching “back”. Similarly, scores are added to headwords that match “nurturing”, “egg”, and “laying”. The headword “reproduction” having the highest total score is selected, and the hierarchical information “ecology-reproduction” is extracted as the metadata of the content.

FIG. 5 is a flowchart illustrating an example of a process for adding metadata to content by the management server 1 according to the present embodiment.
This example is a processing example when the management server 1 is used in a content holder that wants to give metadata to its own video content.
The content holder stores text information necessary for managing video content. This text information is, for example, program information or caption information. The content operator acquires an ontology from the community site by the management server 1 and assigns metadata to the video content.

  In step S <b> 1, the target specifying unit 11 receives the specification of the URL of the Web page to be extracted from the content operator. Alternatively, the target specifying unit 11 acquires text data for analyzing the knowledge structure by taking in a file such as CSV. For example, by specifying the URL of a Web page that describes “daphnia” in the community site, text data related to “daphnia” is obtained. Similarly, by designating a web page of an organism (animal) in the same category as “daphnia” such as “sea urchin” and “dragonfly”, data of other organisms are also added to the extracted data.

In step S2, the headline extraction unit 12 extracts headword hierarchy information from the headline structure of the extracted data acquired from the community site.
In step S3, the word extracting unit 13 extracts a word group characterizing the headword from the word group used in the headline extracted in step S2.

In step S4, the structuring unit 14 generates an ontology that is knowledge structure data based on the hierarchy information of the headword extracted in step S2 and the feature word in the headline extracted in step S3.
The content operator can also manually correct the hierarchical information of the headword, the feature word in the headline, or the generated ontology.

  In step S5, the content operator designates the URL of the video content to which metadata is to be added in order to add metadata based on the ontology from the description text (program information, video content, etc.) of the video content. The content specifying unit 16 acquires text data related to the video content using the specified URL. Alternatively, the content specifying unit 16 may acquire text data by capturing a file such as CSV.

  In step S6, the output unit 17 divides the text data acquired in step S5 into morpheme units. For example, the text “Butterfly larva eats egg shell” is broken down into morphemes “Butterfly”, “larva”, “egg”, “shell”, “eat”.

In step S <b> 7, the output unit 17 calculates which classification each word obtained in step S <b> 6 is most frequently used in the ontology, and determines that “a white butterfly larva eats an egg shell”. On the other hand, for example, metadata “ecology”-“food” is extracted from the ontology and output.
The content operator can acquire the output “ecological”-“food” metadata and use it in his system.

  Also, the designated content can be, for example, a web page of a community site. A page creator can acquire metadata by designating a Web page and rearrange the Web page itself. For example, in a web page related to “daphnia”, if there is a description about a form under the heading “feature”, the page creator redefines “feature” as a common headword “form”, The page itself can be rearranged into content based on a common structure.

Further, for example, when the content operator extracts the characteristics of the table of contents structure and the explanation contents from the textbook data having the text in which the headword is hierarchized as a reference when producing the content for school education, the management server 1 Can be used.
The content operator designates a plurality of URLs in which the textbook data is disclosed to the target designation unit 11 and designates all the URLs to the content designation unit 16 to obtain all the ontology of the textbooks generated by the structuring unit 14. be able to.
The knowledge structure obtained here is, for example, a science textbook for first-year junior high school students, a table of contents such as “organisms and cells”, “animal bodies”, “classification”, etc. It is an ontology that defines characteristic words such as “cell”, “split”, and “egg” that characterize the subordinate table of contents. As a result, the content operator can grasp the table of contents and contents necessary as educational materials. Also, items that are not explained in a specific textbook can be identified.

  According to the present embodiment, the management server 1 extracts hierarchical information of headwords and feature words in the headlines from a plurality of document information having a common field, and integrates them based on the degree of similarity. Generate ontology as structural data. Therefore, the management server 1 can generate knowledge structure data that defines a metadata system that can be commonly used by a plurality of contents.

  For example, in the case of broadcast program content, based on a knowledge structure with commonly used headwords such as “cast”, “broadcast year”, “program related book” generated from a plurality of related web pages, By adding content metadata, metadata definitions can be shared. Also, since the knowledge structure that “program related books” information is described in the broadcast program content is extracted, if the video content is not described, the description contents such as whether or not the “program related books” really exist are scrutinized. Is also possible.

  Since the knowledge structure data includes words associated with headwords, the semantic content of each hierarchical structure of headwords is expressed more specifically. As a result, matching with arbitrary text data becomes easy, and appropriate metadata can be easily given to the content.

  Further, since the knowledge structure data includes the degree of association between the headword and the characteristic word, the degree of matching between the content and the headword can be more specifically compared based on this degree of association. As a result, matching with arbitrary text data becomes easy, and appropriate metadata can be easily given to the content.

  Furthermore, since the management server 1 acquires the synonym of the word in the heading using the dictionary data, matching with the content can be appropriately performed based on the conceptual structure.

  As described above, the content holder can automatically give structured metadata to text information of the content held by using the ontology extracted from the community site as a knowledge source. As a result, the content operator can grasp the content that is duplicated or the content that is lacking in the content that the content operator owns. In addition, the service provider can provide a service such as complementary cooperation between a Web page and video content.

[Second Embodiment]
Hereinafter, a second embodiment of the present invention will be described.
The content server 2 according to the present embodiment is an information processing apparatus (computer) that manages content in a content holder or a service provider. The content server 2 functions as a content acquisition device that acquires new content related to the metadata provided by the management server 1 of the first embodiment.

FIG. 6 is a diagram showing a functional configuration of the content server 2 according to the present embodiment.
The content server 2 includes a metadata acquisition unit 21, a first content acquisition unit 22, and a second content acquisition unit 23.

  The metadata acquisition unit 21 provides its own content (for example, Web page data) to the management server 1 and acquires metadata based on the ontology.

  The 1st content acquisition part 22 acquires the content to which the same metadata as the acquired metadata was provided from a predetermined database (content holder).

  The second content acquisition unit 23 extracts, from the ontology, other hierarchical information having a common upper level of the hierarchical information based on the hierarchical information of the metadata, and the content to which metadata corresponding to the different hierarchical information is given Are obtained from a predetermined database.

FIG. 7 is a diagram illustrating an example of a related content acquisition process according to the present embodiment.
This example shows processing in a service provider who wants to add video content to his Web page.
When a service provider wants to link video content related to the contents of a page to his / her Web page, the service provider can select related content by using metadata provided by the content holder.

  For example, if there are video contents explaining the cultural background of Mt. Fuji and video contents explaining the climate of Mt. Fuji in the content holder, these contents include “mountain”-“faith” and “mountain”-“geology”. It is assumed that the metadata is given.

  When the service provider creates an introduction site for Mt. Fuji, the service provider inquires the management server 1 for a text area 31 describing its cultural background, and acquires metadata “mountain”-“faith”.

  The service provider can correct the headword 32 based on the acquired metadata, search for content related to “mountain”-“faith” possessed by the content holder, and create a link 33. At this time, text 34 accompanying the searched content may be added.

Furthermore, the service provider can grasp that there is video content regarding “mountain”-“geology” by following the conceptual structure of the metadata. Therefore, the service provider can further search and display not only video content related to his / her web page but also systematically related video content. That is, a headword 35 of “geology”, a video link 36 and text 37 accompanying the video are added.
The query for searching for content may be API or SPARQL, but is not limited thereto.

According to the present embodiment, the content server 2 can provide information in cooperation with a plurality of contents by using an ontology by acquiring content with common metadata from the database.
Furthermore, since the content server 2 acquires related content having a common upper hierarchy of metadata based on the ontology, the content server 2 can efficiently collect a plurality of contents including related information and increase the amount of information.

In this way, the content server 2 can not only organize the content by adding metadata to the content based on the ontology using the community site as a knowledge source, but also create new content that complementarily cooperates with other content. Can be produced.
The service provider can easily provide a new service in which a plurality of contents such as a video encyclopedia are linked by using metadata and systematized contents that are shared by using knowledge of the community site. In addition, for example, flexible content linkage such as video content and Web content and video content and video content is possible.

  As mentioned above, although embodiment of this invention was described, this invention is not restricted to embodiment mentioned above. Further, the effects described in the present embodiment are merely a list of the most preferable effects resulting from the present invention, and the effects of the present invention are not limited to those described in the present embodiment.

  In the above-described embodiment, the method for assigning metadata has been described using video content or a Web page as an example. However, the content is not limited to these, and various types of content to which text data that can be matched with an ontology are assigned. And can.

  The functions of the management server 1 (ontology generation device, metadata output device) and the content server 2 (content acquisition device) described above may be provided as a system that is appropriately distributed or integrated depending on the service form.

  In the present embodiment, the configuration and operation of the ontology generation device, and the metadata output device and content acquisition device that use this ontology have been described. However, the present invention is not limited to this, and includes each component to generate an ontology. Alternatively, it may be configured as a method or program for use.

  Furthermore, the program for realizing the function of each device may be recorded on a computer-readable recording medium, and the program recorded on the recording medium may be read by a computer system and executed.

  The “computer system” here includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a hard disk built in the computer system.

  Furthermore, “computer-readable recording medium” means that a program is dynamically held for a short time, like a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line. It is also possible to include one that holds a program for a certain time, such as a volatile memory inside a computer system that becomes a server or client in that case. Further, the program may be for realizing a part of the above-described functions, and may be capable of realizing the above-described functions in combination with a program already recorded in the computer system. .

1 Management server (Ontology generation device, metadata output device)
2 Content server (content acquisition device)
DESCRIPTION OF SYMBOLS 11 Target specification part 12 Headline extraction part 13 Word extraction part 14 Structured part 15 Dictionary acquisition part 16 Content specification part 17 Output part 21 Metadata acquisition part 22 1st content acquisition part 23 2nd content acquisition part

Claims (9)

A headline extraction unit that extracts hierarchical information of headwords from each of a plurality of document information in a specified field;
A word extraction unit that extracts a word associated with the headword from each of the document information;
Ontology generation comprising: a structuring unit that integrates different headwords into one based on the similarity of the words and generates knowledge structure data of the field including hierarchical information of the integrated headwords apparatus.   The ontology generation apparatus according to claim 1, wherein the structuring unit generates knowledge structure data in which the headword and the word are associated with each other.   The ontology generation device according to claim 2, wherein the structuring unit generates the knowledge structure data including a degree of association between the headword and the word.   The hierarchy of the headword associated with the word by matching the text data related to the content with the word included in the knowledge structure data generated by the ontology generation device according to any one of claims 1 to 3. A metadata output device including an output unit that extracts information from the knowledge structure data and outputs the information as metadata of the content. A dictionary acquisition unit that acquires synonyms of the word based on dictionary data;
The metadata output device according to claim 4, wherein the output unit extracts hierarchical information of the headword associated with the word by matching the text data with the word or the similar word.   A content acquisition apparatus comprising: a first content acquisition unit that acquires, from a predetermined database, content provided with the same metadata as the metadata output by the metadata output device according to claim 4 or 5.   Based on the hierarchical information of the metadata, another hierarchical information having a common upper level of the hierarchical information is extracted from the knowledge structure data, and the content to which the metadata corresponding to the different hierarchical information is assigned is the predetermined content. The content acquisition apparatus according to claim 6, further comprising a second content acquisition unit that acquires from the database. The computer controller
A headline extraction step of extracting hierarchical information of headwords from each of a plurality of pieces of document information in a specified field;
A word extracting step of extracting a word associated with the headword from each of the document information;
An ontology that executes the structuring step of integrating the different headwords into one based on the similarity of the words and generating knowledge structure data of the field including hierarchical information of the integrated headwords Generation method. In the control part of the computer,
A headline extraction step of extracting hierarchical information of headwords from each of a plurality of pieces of document information in a specified field;
A word extracting step of extracting a word associated with the headword from each of the document information;
Structuring step of integrating the different headwords into one based on the similarity of the words and generating knowledge structure data of the field including hierarchical information of the integrated headwords Ontology generation program.

Images