JP2018005633A - Related content extraction device, related content extraction method, and related content extraction program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a device, a method and program capable of presenting content dealing with the same event in a time sequence manner.SOLUTION: A related content extraction device 1 comprises a content acquisition unit 11 for acquiring content, an information extraction unit 12 for extracting specific information about an event contained in the content, including event date/time, a similar content retrieval unit 13 for retrieving similar content on the basis of the specific information and a content recording unit 14 for recording the similar content in association with each other according to the issue date/time in a time sequence manner.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an apparatus, a method, and a program for extracting related content.

In recent years, a curation service has been provided that displays news articles collectively by category or topic. Examples of services include a website that summarizes news articles in categories such as “politics” or “gourmet”, and a website that summarizes news articles related to the more specific “Country A terrorist incident”.
Thus, not only providing news articles to general users individually as events occurring at that time, but also providing them as a certain unit is a service that satisfies the intellectual needs of users.

  As a technique for collecting news articles, for example, Patent Document 1 proposes a technique for generating a summary of related articles by adding a connection of article IDs to each article. This is a technique for easily grasping the whole picture of news by summarizing related articles to which a certain article belongs.

JP-A-2005-250648

However, in the conventional curation service, articles having similar contents are gathered, and therefore, time-series order is often not taken into account. For this reason, it has been difficult for the user to correctly understand a series of news articles that are information having time-series connections.
For example, in order to generate an overall image of a related article using the technique of Patent Document 1, an article ID for identifying whether the article is an original article or a subsequent article is manually set for each article in advance. It is necessary to grant. Even when the period is limited, since a summary is generated not based on the content of news but based on the date of issue, only a partial overview can be obtained.

  An object of the present invention is to provide an apparatus, a method, and a program capable of presenting content handling the same event in time series.

  A related content extraction device according to the present invention includes a content acquisition unit that acquires content, an information extraction unit that extracts specific information including an event date and time related to an event included in the content, and searches for similar content based on the specific information A similar content search unit, and a content recording unit that records the similar content in association with each other in chronological order according to an issue date.

  The information extraction unit may extract the location of the event or a person related to the event as the specific information.

  The information extraction unit may derive the event date and time based on a relative expression with the issue date and time included in the content.

  The related content extraction apparatus may include a content presentation unit that selects a content group associated by the content recording unit based on an input search key and outputs the selected content group in time series. .

  The related content extraction method according to the present invention includes a content acquisition step of acquiring content, an information extraction step of extracting specific information including an event date and time related to an event included in the content, and searching for similar content based on the specific information The computer executes a similar content search step and a content recording step of recording the similar content in association with each other in chronological order according to the issue date and time.

  A related content extraction program according to the present invention includes a content acquisition step of acquiring content, an information extraction step of extracting specific information including an event date and time related to an event included in the content, and searching for similar content based on the specific information A similar content search step, and a content recording step of recording the similar content in association with each other in chronological order according to the issue date and time.

  According to the present invention, content that handles the same event can be presented in time series.

It is a figure which shows the function outline | summary of the related content extraction apparatus which concerns on embodiment. It is a figure which shows the structure of the related content extraction apparatus which concerns on embodiment. It is a figure which illustrates the processing content of the information conversion process which concerns on embodiment. It is a figure which illustrates the 1st usage pattern which concerns on embodiment. It is a figure which illustrates the 2nd usage pattern which concerns on embodiment.

Hereinafter, an example of an embodiment of the present invention will be described.
The related content extraction apparatus 1 according to the present embodiment uses a news article (content) from a news article (content) and related past news articles based on the event date / time, place and person information of the event, and other keywords. Are extracted and rearranged in chronological order. At this time, the related content extraction apparatus 1 also presents the keywords of each news article.

FIG. 1 is a diagram showing a functional outline of the related content extraction apparatus 1 according to the present embodiment.
First, the related content extraction apparatus 1 extracts the issue date / time of an article, a keyword indicating the content, and the event date / time, place, and person of an event targeted by the content from each news content.

Subsequently, the related content extraction apparatus 1 searches for related news content for each news content based on the location of the event, the event date and time, the person and the keyword conveyed by all the registered news content, and similar content Define as
The related content extraction apparatus 1 can arrange similar related contents in time series, provide them to a content holder or a curation service provider, and present them to the user.

  For example, the news regarding “terrorism in city B” issued by the content holder as item 90 on November 13, 2015, as the item 95, and the news regarding “memorial throughout the world” as issued as item 95 on November 18, 2015. It is assumed that the contents of the news regarding “J Treaty Agreement” issued on January 5, 2016 as 123 and the news regarding “capture of terrorist criminal” issued on January 5, 2016 as the item 124 are assumed.

The related content extraction apparatus 1 extracts “event date”, “location”, and “person” of events targeted by each news from these news contents and assigns them as data.
For example, for the item 124 issued on January 5, 2016, the term “event date and time” in November 2015 is based on the phrase “Country terrorist attacks by country A in November of last year ...” , Data such as country A is given as “location”. On the other hand, for item 123, “Event Date” is given in September 2015, “City” is I City, and “Person” is D Minister.

The related content extracting apparatus 1 calculates the relevance level of each news content by associating each news content based on “event date and time”, “location”, “person”, and “keyword”. Determine the sequence order. As a result, the content holder or the curation service provider can present the news articles that are the basis of a certain news article in time series.
For example, in the item 124, the item 90 and the item 90 are presented in time series as the news article that is the source.

FIG. 2 is a diagram illustrating a configuration of the related content extraction apparatus 1 according to the present embodiment.
The related content extraction device 1 is an information processing device (computer) such as a server or a PC (Personal Computer), and includes an interface such as an input / output device and a communication unit in addition to a control unit such as a CPU and a storage unit. Good. The related content extraction apparatus 1 implements various functions of the present embodiment by causing the control unit to execute predetermined software (related content extraction program) stored in the storage unit.
The control unit of the related content extraction apparatus 1 includes a content acquisition unit 11, an information extraction unit 12, a similar content search unit 13, a content recording unit 14, and an output unit 15 (content presentation unit and content information output unit). Prepare.

The content acquisition unit 11 acquires the designated processing target content.
The information extraction unit 12 extracts “event date / time”, “location”, “person”, and “keyword” from the acquired content as specific information of the event targeted by the content.

  Here, it is assumed that the content acquired by the content acquisition unit 11 is given the date and time when the content was issued and the text that is the body of the news article. From the content acquired by the content acquisition unit 11, “event date”, “location”, “person”, and “keyword” are extracted by the information conversion process of the information extraction unit 12.

FIG. 3 is a diagram illustrating the processing content of the information conversion processing according to the present embodiment.
First, the information extraction unit 12 decomposes the text of the content into words by morphological analysis, and distributes them into “event date”, “location”, “person”, and “keyword”.

In the case of “event date / time”, the information extraction unit 12 compares the description content in the text with the issue date / time of the content itself, and expresses it in a date / time format that can be expressed absolutely.
For example, in the example of FIG. 3, the date of issue is January 5, 2016, and the content of the text is “Last Year”. In this case, the year indicated by “Last Year” is calculated by subtracting one year from the date of issue, and converted to “2015”. For example, as a rule for subtracting one year, not only “Last year” but also “Last year” is conceivable, and information conversion processing according to the description meaning such date and time is defined in advance by the rule.

The accuracy of the converted date and time can be grasped by the date and time notation after conversion, and is used for searching for similar contents. For example, in the example of FIG. 3, since the extracted “event date” information is November 2015, the search target is limited to events from November 1, 2015 to November 31, 2015. .
On the other hand, for example, when the “event date and time” information is specified as November 13, 2015, the search target of similar content is limited to November 13, 2015.

In the case of “location” and “person”, the information extraction unit 12 may specify the location by referring to a publicly disclosed API for specifying the location or a unique dictionary. For example, the information extraction unit 12 determines whether or not the type of a word obtained by morphological analysis or the like is a person with reference to the latest name space such as DBpedia that is structured from Wikipedia (registered trademark). Also, the location can be determined by a library for Geo coding. These identification methods may be appropriately changed depending on the usage form.
These external or internal resources specify, for example, that “President” is “person”, “B city” is “location”, and that the country is “A country”.
Words that are not classified as “event date”, “location”, or “person” are set as “keywords”.

The similar content search unit 13 searches for similar content based on the extracted “event date”, “location”, “person”, and “keyword”.
The similar content search unit 13 first limits the content search target range based on the “event date / time” information, and further calculates the similarity for each “location”, “person”, and “keyword” of the search target content, Content having high similarity is extracted as similar content.

For example, in the example of FIG. 1, since the event date and time indicated by the news article of item 124 is November 2015, it is the content of November 2015, and items 90 and 95 having words with high similarity are similar contents. Extracted as
The similarity calculated for each “location”, “person”, and “keyword” can be, for example, cosine similarity or context similarity, and a threshold value for determining similar content may be set as appropriate. .

The content recording unit 14 arranges the searched similar contents in order of event date and time and issuance date, and adds the acquired specific information and related information indicating the chronological order of similar contents to the content 20 and accumulates them in the storage 20. .
The data structure in the storage 20 is not limited and can be widely used in various database systems and file systems.

The output unit 15 (content presentation unit) receives a selection or search instruction for content for which the context in the time series is to be grasped, and outputs a similar content group in time series based on the related information of the content recorded in the storage 20 To do. At this time, since a target place, a person, and a used keyword are assigned to each content, a difference between the contents may be presented.
Further, the output unit 15 (content information output unit) associates specific information (“event date and time”, “location”, “person”, “keyword”) and related information on similar content in association with the designated content. You may output as metadata.

FIG. 4 is a diagram illustrating a first usage pattern of the related content extraction apparatus 1 according to this embodiment.
The first usage pattern is an example in which a service provider uses the related content extraction device 1 to collect contents of its own company and other companies and provide curation information. Company X develops a service that collectively presents news content related to a specific topic, and Company Y collects the history before the content owned by the company.
Note that the related content extraction apparatus 1 has already taken in content held by content providers such as the content holders A and B, added time-series related information, and accumulated in the storage 20.

  Service provider X wants to summarize content in a time series regarding terrorism in country A. Therefore, the service provider X makes an inquiry to the related content extraction apparatus 1 using a query such as “Country A” and “terrorism”, so that the content obtained by keyword matching is similar to the same event as this content. Content can be acquired together with time-series related information.

For example, in this example, it can be seen that the item 90 is the first news content, and is then connected to the item 95 and then to the item 124.
In addition to the news text body of each item, the assigned event date, place, person, and keyword are also output. Therefore, the service provider X can visualize the flow of information by displaying not only the content curation rearranged in time series but also specific information such as keywords assigned to each content.

The service provider Y wants to collect content related to the history before the news content “F passenger plane discovery” owned by the company and display it as related content. In this case, it is possible to search with a condition not only for the keyword but also for the issue date and time.
Since the “F Passenger Aircraft Discovery” content has an issuance date of January 15, 2016, if the condition “Before January 15, 2016” is specified, the item 101 is displayed, but the issue date is a condition. The item A-01 that is not satisfied is not displayed as a search result.

FIG. 5 is a diagram illustrating a second usage pattern of the related content extraction apparatus 1 according to this embodiment.
The second usage mode is an example in which the content holder uses the related content extraction device 1 to add data related to “event date”, “location”, “person”, and “keyword” to the content of the company.

  When the content holder Z registers the content Z-A01 related to “countermeasures against terrorism in Japan” in the related content extraction apparatus 1, the “event date”, “location”, “person”, and “keyword” are assigned to the storage. Accumulated. Thereafter, the content holder Z can acquire the specific information provided and the similar content associated with the time series by designating information such as an item number uniquely indicating the company content.

In the example of FIG. 5, the content holder Z can acquire “Japan”, “terrorism”, “countermeasure” and “National Police Agency” as keywords, “K Prime Minister” as a person, and “Japan” as a place. For this reason, the content holder Z can give the acquired information to the content Z-A01 as metadata.
As described above, the user of the related content extraction apparatus 1 can not only confirm time-series relevance with other content, but can also give necessary information to the content itself.

According to the present embodiment, the related content extraction apparatus 1 extracts specific information including an event date and time from the content, searches for similar content based on the specific information, and associates them in time series with the content issue date and time.
Therefore, the related content extraction apparatus 1 can accurately identify a target event in content such as a news article, and can automatically present content that handles the same event in time series.

The related content extraction apparatus 1 further extracts the event location and the related person as the specific information, so that the event targeted by the content can be specified with higher accuracy.
In addition, since the specific information is information that is easily stored and used as a search key by the user, the convenience during the search is improved by being given to the content.

  The related content extraction apparatus 1 extracts the event date and time from the content by converting the relative date and time expression described in the text based on the issue date and time. As a result, an accurate date or date range can be specified regardless of the expression in the text.

  The related content extraction apparatus 1 can systematically present a large number of contents not only by arranging the contents in time series but also by clarifying specific information and keywords representing events targeted by the contents. Further, the related content extraction apparatus 1 can add metadata that can easily identify an event to the content, which leads to enhancement of the content and improvement of user service.

  Furthermore, the related content extracting apparatus 1 divides the connection of each content in time series by providing information divided into time series information of news content, information that can be proper nouns such as place and person, and other keywords. It can also be developed on a map. Therefore, new content curation can be realized, and content presentation supporting user understanding can be realized.

  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.

  Each functional unit in the present embodiment may be appropriately configured as a system separated into a plurality of information processing apparatuses.

  A program for realizing the function of the related content extraction apparatus 1 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, a CD-ROM, or a hard disk built in a computer system.

  In addition, “computer-readable recording medium” means that a program is dynamically stored for a short time, like a communication line when a program is transmitted 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. .

DESCRIPTION OF SYMBOLS 1 Related content extraction apparatus 11 Content acquisition part 12 Information extraction part 13 Similar content search part 14 Content recording part 15 Output part (content presentation part, content information output part)
20 storage

Claims (6)

A content acquisition unit for acquiring content;
An information extraction unit that extracts specific information including an event date and time related to an event included in the content;
A similar content search unit that searches for similar content based on the specific information;
A related content extraction apparatus comprising: a content recording unit that records the similar content in association with a time series according to an issue date.   The related content extraction device according to claim 1, wherein the information extraction unit extracts a location of the event or a person related to the event as the specific information.   The related content extraction apparatus according to claim 1, wherein the information extraction unit derives the event date and time based on a relative expression with the issue date and time included in the content.   4. The content presentation unit according to claim 1, further comprising: a content presentation unit that selects a content group associated by the content recording unit based on an input search key and outputs the selected content group in time series. The related content extraction device described in 1. A content acquisition step of acquiring content;
An information extracting step of extracting specific information including an event date and time related to an event included in the content;
A similar content search step of searching for similar content based on the specific information;
A related content extraction method in which a computer executes a content recording step of recording the similar content in association with a time series according to an issue date and time. A content acquisition step of acquiring content;
An information extracting step of extracting specific information including an event date and time related to an event included in the content;
A similar content search step of searching for similar content based on the specific information;
A related content extraction program for causing a computer to execute a content recording step of recording the similar content in association with a time series according to an issue date and time.

Images