JP2013020460A - Summary preparation device and method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prepare a useful summary on the basis of a part that actually interests many users by simple processing.SOLUTION: A summary based on the part that actually interests many users can be prepared, and the use thereof is not limited because the summary can be prepared without depending upon the length of an original sentence and the number of sentences by specifying a quotation part from a link side page including a URL of a page to be the basis of the summary and performing simple processing for preparing the summary because the quotation part actually attracts a quoting person.

Description

  The present invention relates to a language processing technique.

  In recent years, with the spread of web pages (also simply referred to as “pages” in the present application) and the development of processing technologies, technologies for automatically creating summaries such as news articles have appeared. For example, an example of extracting a predetermined number of characters before and after a search keyword as a summary for each page (called “snippet”) displayed in the web search result can be seen, and a document such as a page is extracted from the beginning and a summary is extracted. Simple methods such as

JP 2009-140411 A

  By the way, the conventional technology as described above was created because the summary is created based only on the document or page (hereinafter referred to as the “original text”) on which the summary is based, although the processing is simple and the load is small. As for the summary, there were problems in usability because not all parts of the original text were actually interesting.

  In order to achieve both comprehensiveness and readability as a summary, a similar sentence similar to the sentence in the sentence to be summarized is extracted from the sentence on the web, and the frequency of appearance of this similar sentence and the connection between similar sentences are extracted. A device for creating a summary sentence using the appearance frequency of a pattern has also been proposed (see, for example, Patent Document 1). However, this proposal outputs the result of complicated and mechanical processing based on a plurality of sentences of a certain length, which is not suitable for short original sentences, has limited use, and has a heavy processing load. However, it was not based on the actual interests of the company, but did not solve the problems of the prior art.

  In order to solve the above problems, an object of the present invention is to create a useful summary based on a part that many users are actually interested in by simple processing.

  Based on the above object, the summary creation device according to one aspect (1) of the present invention provides a URL acquisition means for acquiring a URL of a page as a basis for creating a summary, and a character string in the page related to the URL as a text. A source text acquisition unit that acquires as a citation destination page, a citation destination page extraction unit that extracts a page including the acquired URL as a citation destination page, and a citation destination that acquires a character string in the extracted citation destination page A text acquisition means, a collation specifying means for specifying a citation part by comparing the acquired original text and the quoted text, a summary creation means for creating a summary based on the specified citation part, It is characterized by having.

  Another aspect (4) of the present invention is a summary creation method that captures the above aspects in the category of the method, a URL acquisition process for acquiring a URL of a page that is a basis of the summary creation, and the URL related to the URL Source text acquisition processing for acquiring a character string in a page as a source text, citation destination page extraction processing for extracting a page including the acquired URL as a citation destination page, and quoting a character string in the extracted citation destination page A citation destination text acquisition process to be acquired as a destination text, a verification specification process for specifying a citation portion by comparing the acquired original text and the citation destination text, and a summary based on the specified citation portion. The summary creation process to be created is executed by a computer.

  Another aspect (5) of the present invention is a summary creation program that captures the above aspect in the category of a computer program. By controlling the computer, the URL of a page that is the basis of the summary creation is acquired. A character string in the page related to the URL is acquired as a source text, a page including the acquired URL is extracted as a citation destination page, and a character string in the extracted citation destination page is acquired as a citation destination text. The obtained original text and the quoted text are collated to identify a cited part, and a summary is created based on the identified cited part.

  In this way, by performing a simple process of creating a summary by identifying the quoted part from the linked page that contains the URL of the page that is the basis of the summary, the quoted part is the part that the quoter is actually interested in Therefore, it is possible to create a summary based on the part that many users are actually interested in, and it is possible to create a summary without depending on the length of the original text or the number of sentences. Applicable.

  Another aspect (2) of the present invention is characterized in that, in any one of the above aspects, the collation specifying means sets a predetermined amount before and after the URL in the quoted text as the object of collation.

  In this way, by setting a predetermined amount before and after the URL in the quoted text as a target of collation, a text portion that seems to have little relation to the content of the original text located at a predetermined amount away from the URL can be omitted from the target. This makes it possible to reduce the verification load and improve the accuracy of summarization.

  According to another aspect (3) of the present invention, in any one of the aspects described above, the collation specifying means determines the part related to a predetermined reference sign in the reference text as the reference part in preference to the other part. It is characterized by specifying.

  As described above, in the quoted text, the part related to a predetermined quotation mark such as the brackets, quotation marks, and HTML emphasis tag is used in the summary in preference to the other parts such as increasing the score of the selection criterion. This makes it possible to obtain an excellent summary with high accuracy and accuracy, including a more impressive part based on the part highlighted by the quotation mark by the person who wrote the blog.

  It should be noted that categories different from the above-described aspects (methods for apparatuses, programs for methods, etc.) and more specific aspects described below are also included in the present invention. For categories different from the above-described aspects, “means” is appropriately read as “processing” or “step”. Further, the order of execution of processes and steps is not limited to the above, and can be changed as appropriate or can be processed collectively. Furthermore, in the category of methods and programs, “computers” that execute individual processes may be common or may be different for each process.

  According to the present invention, it is possible to create a useful summary based on a portion where many users are actually interested in a simple process.

The functional block diagram which shows the structure of embodiment of this invention. The figure which illustrates the quotation destination page (blog etc.) in embodiment of this invention. The flowchart which shows the process sequence in embodiment of this invention. The conceptual diagram which shows the operation example of embodiment of this invention. The figure which shows the other example of the quotation destination page in embodiment of this invention.

  Next, as an example of the present invention, a mode for carrying out the present invention (referred to as “embodiment”) will be described with reference to the drawings. It should be noted that assumptions common to those already described in the background art and problems are omitted as appropriate.

[1. Constitution〕
The present embodiment relates to a summary creation device 1 (also referred to as “this device” or “this device 1”) shown in FIG. 1 (configuration diagram). It is created by utilizing the quotation destination (for example, FIG. 2). In the present embodiment, it is assumed that the website administrator using the terminal T requests the apparatus 1 to summarize the articles in the news server NS. As shown in FIG. 1, the apparatus 1 includes at least an arithmetic control unit 6 such as a CPU, a storage device 7 such as a main memory and an auxiliary storage device, and a communication network N (for example, the Internet, a mobile phone network). Communication means 8 (communication gateway device, mobile phone network, communication circuit with PHS network, wireless LAN adapter, etc.).

  The terminal T is an electronic information device such as a stationary PC, a mobile PC, a tablet PC, a smartphone, and a mobile phone terminal. In addition to the configuration of the computer as described above, although not illustrated, a liquid crystal display panel, a touch panel, It has an input / output unit using push buttons. The terminal T shown in FIG. 1 is used by a website administrator who creates a summary, and illustration of a terminal used by an end user such as a general website viewer is omitted.

  Further, in the present apparatus 1, elements (20, 30, etc.) such as each means shown in FIG. 1 are realized by a predetermined computer program stored (installed) in the storage device 7 controlling the arithmetic control unit 6. . Among these elements, information storage means can be realized in any format such as various databases (also referred to as “DB”), files, arrays, variables such as various stacks, registers, system setting values in the storage device 7, Although not shown, the material and result of processing by each means are stored.

  In addition, the arrow in a figure shows the main direction supplementarily about flows, such as data and control, and does not mean limitation of a direction. For example, in order to acquire web data in a certain direction, there is a page request transmission in the reverse direction in advance. In addition, each means other than the storage means is a processing means for realizing and executing the following information processing functions / actions, but these are functional units arranged for explanation, actual hardware elements and It doesn't matter if it matches the software module.

[2. Main actions and effects)
The flowchart of FIG. 3 shows a processing procedure in which the present apparatus 1 generates a summary in response to a request in the present embodiment configured as described above. A broken line and a broken line arrow indicate information based on the previous processing step. Represents the usage relationship used in later processing steps. In this processing procedure, first, the URL acquisition means 20 acquires the URL of a page (hereinafter referred to as “original page”) that is a basis for creating a summary (step S2).

  As for the URL to be acquired, for example, in the creation request from the terminal T, the URL of the article on the news server NS that is the subject of the summary creation may be individually specified or set as the creation target in advance. The URL acquisition unit 20 may acquire from the news server NS for each article within a predetermined range (for example, in a predetermined directory). Other than these, any acquisition mode may be used, such as automatically detecting the URL of an article that satisfies some condition (such as the number of browsing more than a predetermined number). In the present application, “URL” is not limited to a narrowly defined URL, but means identification information for specifying a page in a data group such as WWW (World Wide Web), and may be an IP address or the like.

  Subsequently, the original text acquisition means 30 acquires a character string (for example, the title and body of a news article) in the original page related to the URL acquired as described above as the original text from the news server NS (step S3). . Further, the citation destination page extraction unit 40 typically extracts a page including the acquired URL as a citation destination page from the web (step S4). An example of a cited page is a web page representing a blog article or comment provided by the blog server BS or the like, a post message of a mini blog (so-called “tweet”, etc.), etc., but a page such as a commercial article including its URL But you can.

  Further, the “page” is not limited to individual pages such as pages DA, DB, and DC illustrated in FIG. 4, but as illustrated in FIG. 5, a single web page such as an electronic bulletin board or a miniblog. Individual elements when a plurality of elements such as articles and comments are included in P, that is, page portions may be used. In the example of FIG. 5, there are articles posted by a plurality of users such as Mr. A, Mr. B, Mr. C, and Mr. D on the web page P, and among them, an article C1 including the URL of the article on the original page Only C3 (indicated by a broken line ellipse) is extracted as a cited page.

  In addition, although the method of extracting these citation pages can be freely selected, it is typically considered that a web search request is transmitted to the search server SS using the URL as a search key and the search result is used. Then, the citation destination text acquisition unit 50 acquires the extracted character string in the citation destination page as the citation destination text (step S5). For the text to be extracted, except for the typical elements (advertising column, company profile, links to help, etc.) of pages such as blog sites, the net part such as blog articles and comments can be selectively extracted. Increases accuracy.

  And the collation specific | specification means 60 identifies a quotation part by collating the acquired original text and quotation text (step S6). The criteria for specifying the citation part are also free, but typically, a part that is a character string in the original text and is included in the citation destination text of a predetermined ratio or more is specified as the citation part. For example, in the example of FIG. 4, the character string “television television” included in the original text in the original page D1 (indicated by a dashed ellipse in the figure) is the URL of the original page D1 (indicated by an underline in the broken line in the figure). ) Including the quotation destination pages DA, DB, and DC (indicated by a dashed ellipse at the end of the broken arrow in the figure), it can be specified as a quotation portion.

  Moreover, the summary creation means 70 creates a summary based on the identified citation part (step S7). "Based on the quoted part" means that the identified quoted part can be used as a summary (indicated by a broken-line rectangle in the figure) as shown by a one-dot chain line in FIG. 4, or a plurality of identified quoted parts It may be processed such as by joining together, and it also includes the adoption of a large amount of matching parts with the original text (for example, the number of consecutively matched characters and bytes). . The “creation” of the summary may be a process of outputting the quoted part as a summary as it is to an output device, a subsequent process, a process or a routine that is a function caller, or the like.

  As mentioned above, the citation was actually attracted by the citation by identifying the citation from the linked page that contains the URL of the page that is the basis of the summary and creating a summary. Since it is a part, it is possible to create a summary based on a part that many users are actually interested in, and to create a summary regardless of the length of the original text or the number of sentences.

[3. (Specification of quoted part)
In particular, it is desirable that the collation specifying unit 60 uses a predetermined amount before and after the URL in the quoted text as a collation target. For example, in the blog article illustrated in FIG. 2, the 100 characters immediately before and 100 characters immediately after the URL of the cited article (indicated by a dashed ellipse in the figure) are TV topics related to the article content. Even before and after, it has become an irrelevant topic such as stray cat measures and alumni associations.

  As described above, since the content related to the cited URL is generally described immediately before or after the URL, by using a predetermined amount before and after the URL in the quoted text, A text portion that seems to have little relation to the content of the original text located at a predetermined distance from the URL can be omitted from the target, reducing the collation load and improving the accuracy of the summary. Of course, the predetermined amount before and after the URL is not limited to 100 characters, and may be freely determined such as the number of other characters, the number of lines, the number of bytes of data, and even blank lines. It may be at least one immediately before or after, or the amount may be different between immediately before and after (for example, 100 characters immediately before, 200 characters immediately after, etc.).

  As illustrated in FIG. 5, when the same URL exists at a plurality of locations in a single page, the predetermined amount is set as a target of collation based on each URL. In this case, the “predetermined amount” of the character string to be collated may be the same as the case where there is only one URL in a single page (100 characters in the above example), but a smaller amount (for example, 20 Character). If you do this, a part of multiple posts will be included in a single page as excerpts, so if you shift 100 characters back and forth, it will become another post, or if the number of individual post characters is limited in the first place (For example, 140 characters, etc.) Even if the topic is a post or page structure that switches in a short span, the text portion that seems to have little relation to the original text can be reliably excluded from the target, so the summary is excellent Accuracy can be maintained.

  Further, the collation specifying means 60 may increase the selection criterion score for a portion related to a predetermined quotation mark (for example, brackets, quotation marks, HTML emphasis tag, etc.) in the citation destination text. It is desirable to specify it as a cited part with higher priority and use it in the summary. For example, in the example of FIG. 4, in the cited page DA, the quote part that quotes the original text is attached with the bracket of “behind the television” and similarly, the cited page DB is quoted at the beginning of the cited part. A double quote "" "which is a quotation mark in the quotation destination page DC is added to the quotation mark" >> ".

  For the parts related to such quotation marks, the criteria and mode for specifying the quotation part in preference to the other parts are free, but in the original text, the part included in more than 10% of the cited page is referred to as the quotation part. In the case of the selection criterion such as “Yes”, the calculation load is reduced if the portion related to the quotation mark is counted with a weight increased from that of a normal case such as two or three cases.

  In this way, the part related to a given quotation mark in the cited text is identified as the quotation part in preference to the other parts, and used for summarization, so that the person who wrote the blog etc. emphasized with the quotation mark in particular Based on the above, an excellent summary with high precision and accuracy can be obtained.

[4. Other embodiments]
In addition, the said embodiment is only an illustration and this invention includes what is illustrated below and other embodiment other than that. For example, the configuration diagrams, data diagrams, flowcharts, and the like in the present application are merely examples, and the presence / absence of each element, the order of the arrangement and processing execution, and the specific contents can be changed as appropriate.

  As an example, in the above embodiment, an SNS (social networking service) page such as a blog stored in a blog server BS or the like is exemplified on the Internet as a reference page that is a basis for creating a summary. As for the cited page, a page on the intranet or social media represented by a simple electronic bulletin board is free.

  The summary created in the above embodiment can be used by a website administrator using the terminal T, for example, to display a list of topics of news websites provided by the news server NS. Applications are not limited to such operations. For example, a general end user accesses the apparatus 1 from a browser of a personal computer (PC) to use the summary according to the present invention or installs a program according to the present invention on a PC in order to grasp the outline of each article. You may get a summary.

  In addition, a configuration in which the individual means shown in FIG. 1 and the like are realized by separate and independent facilities is also common, and depending on the function of the server or the terminal, an external platform or the like may be connected to an API (Application Program Interface). ) And network computing (so-called cloud etc.), and the configuration can be flexibly changed. For example, the apparatus and each server shown in FIG. 1 may be integrated as appropriate. Furthermore, each element such as means relating to the present invention may be realized by other information processing mechanisms such as a physical electronic circuit as well as a computer control unit.

1 Summary creation device (this device)
6 Arithmetic Control Unit 7 Storage Device 8 Communication Unit 20 URL Acquisition Unit 30 Original Text Acquisition Unit 40 Citation Destination Page Extraction Unit 50 Citation Destination Text Acquisition Unit 60 Collation Identification Unit 70 Summary Creation Unit BS Blog Server D1 Original Page DA, DB, DC Citation Previous page N Communication network NS News server SS Search server T Terminal

Claims (5)

URL acquisition means for acquiring the URL of the page that is the basis for the summary creation;
Original text acquisition means for acquiring a character string in the page related to the URL as the original text;
Citation page extraction means for extracting a page including the acquired URL as a citation page;
A quoted text acquisition means for acquiring a character string in the extracted quoted page as a quoted text;
A collation specifying means for identifying a citation part by collating the acquired original text with the quoted text;
A summary creation means for creating a summary based on the identified cited part;
A summary creation device characterized by comprising:   2. The summary creation device according to claim 1, wherein the collation specifying unit uses a predetermined amount before and after the URL in the quoted text as a target of collation.   3. The summary creation device according to claim 1, wherein the collation specifying unit specifies a portion related to a predetermined reference code in the reference text as the reference portion in preference to other portions. URL acquisition processing for acquiring the URL of the page that is the basis for creating the summary;
An original text acquisition process for acquiring a character string in the page related to the URL as an original text;
A citation page extraction process for extracting a page including the acquired URL as a citation page;
A quoted text acquisition process for acquiring a character string in the extracted quoted page as a quoted text;
A collation specifying process for identifying a citation part by collating the acquired original text with the quoted text;
A summary creation process for creating a summary based on the identified cited part;
A method for creating a summary, characterized in that a computer executes. By controlling the computer
Get the URL of the page that is the basis of the summary creation,
The character string in the page related to the URL is acquired as the original text,
The page including the acquired URL is extracted as a citation page,
The extracted character string in the cited page is acquired as the cited text,
By collating the acquired original text and the cited text, the citation part is specified,
A summary creation program for creating a summary based on the identified cited part.

Images