JP2010286997A - Device and method for evaluating value of article, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a device and method for automatically evaluating the value of an article, and for extracting a useful article, and a program. <P>SOLUTION: An evaluation server 10 is provided with: a pattern DB22 for storing one or more normalized pattern configured by including the combination of items included in the past article; an article pattern classification part 13 for classifying an evaluation object article and the past article into any of the normalized patterns; a word/phrase extraction part 14 for extracting words/phrases in the evaluation object article and the past article corresponding to one or more items included in the classified normalized pattern; an appearance frequency DB23 for storing statistical data associating the appearance frequencies for each of the extracted words and phrases; and a value evaluation part 15 for calculating an evaluation value showing the value of the evaluation object article on the basis of the deviation of the appearance frequency of the words and phrases included in the evaluation object article, in the stored statistical data. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to an apparatus, a method, and a program for evaluating the value of an article.

  2. Description of the Related Art Conventionally, a service has been provided in which a large number of news articles transmitted from newspaper companies, communication companies, etc. are selected and posted on Internet Web pages. In such a service, it is desirable to display a concise and easy-to-read list of articles that are interesting and useful for many users, that is, high-value articles.

  Therefore, a technique for automatically generating a summary sentence from an article has been proposed. For example, Patent Document 1 discloses an apparatus that extracts word information from an article by collating pattern information with an article and generates a summary sentence using a template.

Japanese Patent Laid-Open No. 2002-197097

  However, although the apparatus of Patent Literature 1 may be used for list display by generating a summary sentence of articles, it extracts useful articles to be provided to the user from a large number of articles. That was difficult.

  An object of the present invention is to provide an apparatus, a method, and a program capable of automatically evaluating the value of an article and extracting a useful article.

  The present invention provides the following solutions.

(1) A device for evaluating the value of an article,
Pattern storage means for storing one or more normalized patterns that include a combination of items included in past articles;
Classification means for classifying the article to be evaluated and the past article into any of the normalized patterns stored by the pattern storage means;
Extraction means for extracting words in the evaluation target article and the past article corresponding to one or more items included in the normalized pattern classified by the classification means;
For each of the phrases extracted by the extraction means, an appearance count storage means for storing statistical data associated with the appearance count in the evaluation target article and the past article;
In the statistical data stored by the appearance count storage means, a calculation means for calculating an evaluation value indicating the value of the evaluation target article based on a deviation of the appearance count of the words included in the evaluation target article. apparatus.

  According to such a configuration, the apparatus stores one or more normalized patterns that include a combination of items included in past articles, stores the evaluation target article and the past article, Classification into any one of the normalization patterns, and words in the evaluation target article and the past article corresponding to one or more items included in the classified normalization pattern are extracted. Further, the apparatus stores statistical data associated with the number of appearances in the evaluation target article and the past article for each of the extracted words and phrases, and in the stored statistical data, the phrase included in the evaluation target article is stored. An evaluation value indicating the value of the evaluation target article is calculated based on the deviation of the number of appearances.

  Therefore, the apparatus can extract a phrase for each item by collating with the normalization pattern. Then, the apparatus calculates the evaluation value of the article to be evaluated based on the deviation of the number of appearances based on the relationship between the number of appearances and the number of phrases obtained by performing statistical processing on the extracted words, specifically, the deviation value. be able to. For example, since there is a high possibility that the number of appearances is an important word that should be noticed as the number of appearances deviates from the average, the degree of importance can be measured by the deviation value. As a result, the device may automatically evaluate the value of the entire evaluation target article based on the importance of the phrase for each item, and extract a useful article.

(2) further comprising value criterion accepting means for accepting an input of a value criterion that defines the importance of the phrase with respect to the deviation of the number of appearances;
The apparatus according to (1), wherein the calculation unit calculates the evaluation value by combining the importance of words for each item included in the article to be evaluated.

  According to such a configuration, the apparatus receives an input of a value standard that defines the importance of the phrase with respect to the deviation in the number of appearances, and evaluates the importance of the phrase for each item included in the article to be evaluated. Is calculated.

  Therefore, the device can calculate the importance of the phrase from the deviation of the number of appearances based on the value criterion for which the input has been received in advance. Thereby, the apparatus can calculate the value of the evaluation target article as an evaluation value that is a specific index by integrating the importance of words for each item. Note that an evaluation value obtained by combining the importance levels of words can be calculated by a predetermined calculation such as a total or average of the importance levels.

(3) selection means for comparing the importance of the words for each item included in the article to be evaluated, and selecting a word with a relatively high importance;
(2) The apparatus according to (2), further comprising generation means for generating title data composed of the words selected by the selection means.

  According to such a configuration, the apparatus compares the importance of words for each item included in the evaluation target article, selects a word with a relatively high importance, and a title including the selected word Generate data.

  Therefore, the device can select a word that is highly important and attracts attention from the user. And since the said apparatus comprises a title by such attention words and phrases, the notification effect can be improved especially on conditions, such as the number of characters being limited.

(4) The selection means selects a predetermined amount of words / phrases set in advance,
The apparatus according to (3), wherein the generation unit generates the title data by arranging the words selected by the selection unit in order of the importance.

  According to such a configuration, the apparatus selects a predetermined amount of words / phrases set in advance, and generates the title data by arranging the selected words / phrases in descending order of importance.

  Therefore, the device can preferentially display words with higher importance as titles. As a result, the degree of attention of the title increases, and an improvement in notification effect can be expected.

  (5) The apparatus according to any one of (2) to (4), wherein the value criterion accepting unit accepts an input of weighting of the importance level that is different for each item as the value criterion.

  According to such a configuration, the device can set different importance weights for each item based on the input of the value criterion from the administrator. Therefore, the apparatus can adjust the value criteria flexibly and appropriately set the type of article that has high value.

  (6) In any one of (2) to (5), the value criterion accepting unit accepts, as the value criterion, an input of the weighting of the importance level that differs for each acquisition source of the evaluation target article. The device described.

  According to such a configuration, the apparatus can set different importance weights for each acquisition source of the evaluation target article, that is, for each newspaper company, news agency, etc., based on the input of the value criterion from the administrator. it can. Therefore, the apparatus can set the reliability and the importance according to the source of the article, and appropriately set the type of the article that has a high value.

  (7) The apparatus according to any one of (2) to (6), wherein the value criterion accepting unit accepts, as the value criterion, an input of weighting of the importance level that is different for each normalization pattern. .

  According to such a configuration, the apparatus can set different importance weights for each normalization pattern based on the input of the value criterion from the administrator. Therefore, the apparatus can preliminarily extract important article types (normalization patterns) in advance.

(8) rank criterion receiving means for receiving an input of a rank criterion that defines the priority of a plurality of words corresponding to the same item;
The apparatus according to any one of (1) to (7), further comprising: a determination unit that determines the rank of a plurality of articles having the same evaluation value based on the rank criterion.

  According to such a configuration, the apparatus receives an input of a rank criterion that defines the priority of a plurality of words corresponding to the same item, and based on the rank criterion, ranks a plurality of articles having the same evaluation value. to decide.

  Therefore, the apparatus can rank a plurality of articles having the same evaluation value based on the ranking criteria input from the administrator. That is, the apparatus can preferentially extract articles including words with high priority.

(9) A method by which a computer evaluates the value of an article,
A pattern storage step of storing one or more normalized patterns that are configured to include combinations of items included in past articles;
A classification step of classifying the article to be evaluated and the past article into any of the normalized patterns stored in the pattern storage step;
An extraction step of extracting words in the evaluation target article and the past article corresponding to one or more items included in the normalized pattern classified by the classification step;
For each of the phrases extracted in the extraction step, an appearance count storage step for storing statistical data associated with the appearance count in the evaluation target article and the past article;
In the statistical data stored in the appearance count storage step, a calculation step of calculating an evaluation value indicating the value of the evaluation target article based on a deviation of the appearance count of the words included in the evaluation target article is included. Method.

  According to such a configuration, the same effect as in (1) can be expected by executing the method.

  (10) A program that causes a computer to execute the method according to (9).

  According to such a configuration, the same effect as in (1) can be expected by causing the computer to execute the program.

  According to the present invention, it is possible to automatically evaluate the value of articles and extract useful articles.

1 is a diagram illustrating an overall configuration of a system according to an embodiment of the present invention. It is a figure which shows the hardware constitutions of the evaluation server which concerns on embodiment of this invention. It is a figure which shows the function structure of the evaluation server which concerns on embodiment of this invention. It is a figure which shows an example of the normalization pattern which concerns on embodiment of this invention. It is a figure which shows an example of the statistical data which concern on embodiment of this invention. It is the figure which showed typically the shape about an example of distribution of the appearance frequency which concerns on embodiment of this invention. It is a figure which shows an example of the value reference | standard table which concerns on embodiment of this invention. It is a flowchart which shows the process example which concerns on embodiment of this invention.

  Hereinafter, an example of an embodiment of the present invention will be described with reference to the drawings.

[Entire system configuration]
FIG. 1 is a diagram showing an overall configuration of a system including an evaluation server 10 and related elements according to the present embodiment. In this system, an evaluation server 10, a plurality of content servers 20 that transmit article data, a Web server 30, and a user terminal 40 are connected to each other via a predetermined network such as the Internet.

  The evaluation server 10 receives news articles that are distribution candidates from the content server 20. The evaluation server 10 generates a normalization pattern of past articles by statistical processing, and stores statistical data related to the number of appearances of words for each item. Then, the evaluation server 10 refers to this statistical data, and evaluates the value of the received article based on the importance corresponding to the deviation value related to the number of occurrences of the phrase.

  When an article with a high evaluation value is selected from the evaluation server 10 and transmitted to the Web server 30, the Web server 30 displays a list of articles (topics display) in a predetermined area of the Web page. Thereby, when the user terminal 40 accesses the Web server 30 to display a predetermined Web page, the user can browse the article of interest automatically extracted by the evaluation server 10.

[Hardware configuration]
FIG. 2 is a diagram illustrating a hardware configuration of the evaluation server 10 according to the present embodiment. The evaluation server 10 includes a CPU (Central Processing Unit) 310 (a plurality of CPUs such as a CPU 320 may be added in a multiprocessor configuration), a bus line 200, and a communication I / F (I / F: interface) that configure the control unit 300. ) 330, a main memory 340, a BIOS (Basic Input Output System) 350, an I / O controller 360, a hard disk 370, an optical disk drive 380, and a semiconductor memory 390. The hard disk 370, the optical disk drive 380, and the semiconductor memory 390 are collectively referred to as a storage device 410.

  The control unit 300 is a part that controls the evaluation server 10 in an integrated manner. By appropriately reading and executing various programs stored in the hard disk 370 (described later), the control unit 300 cooperates with the hardware described above, and Various functions are realized.

  The communication I / F 330 is a network adapter when the evaluation server 10 transmits / receives information to / from the content server 20, the web server 30, and other information terminals in FIG. 1 via the network. The communication I / F 330 may include a modem, a cable modem, and an Ethernet (registered trademark) adapter.

  The BIOS 350 records a boot program executed by the CPU 310 when the evaluation server 10 is started up, a program depending on the hardware of the evaluation server 10, and the like.

  A storage device 410 such as a hard disk 370, an optical disk drive 380, and a semiconductor memory 390 can be connected to the I / O controller 360.

  The hard disk 370 stores various programs for causing the hardware to function as the evaluation server 10 and programs for executing the functions of the present invention. Note that the evaluation server 10 can also use an external hard disk (not shown) as an external storage device.

  As the optical disk drive 380, for example, a DVD-ROM drive, a CD-ROM drive, a DVD-RAM drive, or a CD-RAM drive can be used. In this case, the optical disk 400 corresponding to each drive is used. A program or data can be read from the optical disk 400 by the optical disk drive 380 and provided to the main memory 340 or the hard disk 370 via the I / O controller 360.

  The computer referred to in the present invention refers to an information processing device including a storage device, a control unit, and the like. The evaluation server 10 includes an information processing device including a storage device 410, a control unit 300, and the like. The processing device is included in the concept of the computer of the present invention.

[Function configuration]
FIG. 3 is a diagram illustrating a functional configuration of the evaluation server 10 according to the present embodiment. The control unit 300 of the evaluation server 10 includes an article pattern generation unit 11, an article reception unit 12, an article pattern classification unit 13 (classification unit), a phrase extraction unit 14 (extraction unit), and a value evaluation unit 15 (calculation unit). Determination means), attention phrase selection unit 16 (selection unit), title generation unit 17 (generation unit), value standard reception unit 18 (value standard reception unit), and rank reference reception unit 19 (rank standard reception unit). And). The storage device 410 of the evaluation server 10 includes an article DB 21, a pattern DB 22 (pattern storage unit), an appearance count DB 23 (appearance count storage unit), and an evaluation criterion DB 24.

  Hereinafter, each functional unit will be described as being included in the evaluation server 10, but the configuration is not limited thereto. For example, each functional unit may be distributed to a plurality of servers. Moreover, you may integrate with other servers, such as Web server 30. FIG.

  The article pattern generation unit 11 normalizes past article data stored in the article DB 21 with a combination of evaluation target items, and generates a normalized pattern. In addition, the above-mentioned existing technique etc. can be utilized for a production | generation method. Thereby, for example, a plurality of normalization patterns such as “traffic accident”, “finance”, “weather forecast”, “sports”, etc. can be stored.

FIG. 4 is a diagram illustrating an example of a normalization pattern according to the present embodiment.
In this example, the evaluation target items included in the normalization pattern “traffic accident 1” are shown. Each article data has one or more words / phrases associated with the evaluation target item. In this example, for example, the phrase of the evaluation target item “party-age” is “102 years old”, and the phrase of the evaluation target item “search content—profession” is “professional baseball player”.

  The article receiving unit 12 receives an input of an evaluation target article. In addition, although this evaluation object article is demonstrated as what is received from the content server 20 as a new article, it is not restricted to this. For example, you may select from article DB21.

  The article pattern classification unit 13 classifies the evaluation target article received by the article reception unit 12 into one of the normalization patterns. Specifically, the evaluation target article is collated with the pattern DB 22 and classified into a normalized pattern having the highest degree of matching or similarity.

  The phrase extraction unit 14 extracts a phrase in the corresponding evaluation target article for each of the evaluation target items included in the normalized pattern classified by the article pattern classification unit 13.

  Moreover, the phrase extraction part 14 memorize | stores the frequency | count of appearance of a corresponding phrase as statistical data in appearance frequency DB23 for every evaluation object item shown in FIG. At this time, the article that is the target of statistical data may be an article of a predetermined amount or a predetermined period that is set in advance, including past articles stored in the article DB 21.

  FIG. 5 is a diagram illustrating an example of statistical data stored in the appearance count DB 23 according to the present embodiment. In this example, the evaluation target item is “age”, the number of appearances is plotted on the horizontal axis, and the number of phrases (age numbers) is plotted on the vertical axis.

  In the statistical data, the number of words having the same appearance number is associated with the number of appearances for each of the evaluation target items. For example, the ages “2 years old” and “101 years old” each appear only once in the articles targeted for statistics, so the number of appearances is recorded as “1 time”, and the number of appearances is “1 time”. The number of phrases “” is “2”. Similarly, the number of words with the number of appearances “50” is “6”, which is larger than the number of other appearances.

  In this way, the number of words is accumulated for each number of appearances, and a distribution of the number of appearances is formed for each evaluation target item.

FIG. 6 is a diagram schematically showing the shape of an example of the appearance frequency distribution according to the present embodiment.
In this example, a portion with a small number of words / phrases having the same number of appearances has a large deviation value from the average and is likely to be a special word / phrase. That is, such a special word / phrase is a word / phrase with high importance that characterizes the article, and further, the value of the article is considered high.

  The shape of the distribution of the number of appearances differs for each evaluation target item, but the importance of the phrase can be relatively determined using the deviation value of the number of appearances as an index. That is, in the vicinity of the number of appearances where the deviation value is far from the average, the number of phrases is small, and the importance of these phrases is considered high.

  The value evaluation unit 15 refers to the appearance count DB 23 and the evaluation criterion DB 24 for the phrase extracted by the phrase extraction unit 14 and calculates the importance of the phrase. Further, the value evaluation unit 15 combines the importance levels of a plurality of words and calculates an evaluation value indicating the value of the entire evaluation target article. Specifically, the evaluation value is calculated by performing predetermined statistical processing such as averaging and summing on the importance of the phrase.

  Here, the evaluation criterion DB 24 is a DB that defines the importance of the word with respect to the deviation value of the number of appearances of the word obtained by the appearance number DB 23.

FIG. 7 is a diagram illustrating an example of the value criterion table stored in the evaluation criterion DB 24 according to the present embodiment.
In this example, the deviation value and the importance are stored in association with each evaluation target item. Specifically, the importance of the word is set higher as the deviation value is smaller than the average (deviation value = 50) or larger than the average. As a result, the importance of words with a frequency of appearance that is significantly different from the usual can be increased and reflected in the evaluation of the value of the evaluation target article.

  Note that the relationship between the deviation value and the importance is not limited to this, and may be appropriately set according to the distribution shape of the number of appearances.

  The attention phrase selecting unit 16 selects a phrase having a relatively high importance from the phrases extracted by the phrase extracting unit 14. At this time, it is assumed that selection is made from the highest importance within a predetermined amount, for example, a predetermined number of characters.

  The title generation unit 17 generates title data including words with high importance selected by the attention word / phrase selection unit 16. Specifically, the title data may be concatenated with delimiters (for example, blanks) in descending order of importance. At this time, since a predetermined amount of words / phrases are extracted by the attention word / phrase selector 16, title data can be generated efficiently when there is a restriction on the number of characters or the like.

  The value criterion receiving unit 18 receives an input of a value criterion from the administrator of the evaluation server 10 and stores it in the evaluation criterion DB 24. In addition to the above-described value standard table (FIG. 7) in which the deviation value and the importance are associated, the value standard is different for each evaluation target item, different weights for each acquisition source of the evaluation target article, and is different for each normalization pattern. Weighting etc. may be included. The value reference receiving unit 18 receives and stores these weighted inputs, thereby flexibly adjusting the importance of the word and the value of the article to be evaluated calculated by the value evaluating unit 15 to become high value. Article types can be set as appropriate.

  The ranking reference receiving unit 19 receives an input of ranking criteria from the administrator of the evaluation server 10 and stores it in the evaluation criteria DB 24. The rank criterion specifies the priority of a plurality of words / phrases corresponding to the evaluation target item. When the evaluation values of the plurality of articles to be evaluated are equal, the value evaluation unit 15 determines the priority order based on the rank criterion.

[Processing flow]
FIG. 8 is a flowchart illustrating a processing example of the control unit 300 in the evaluation server 10 according to the present embodiment. In this example, the control unit 300 selects and presents a high-value article from among a plurality of evaluation target articles. It is assumed that the normalization pattern of the pattern DB 22, the statistical data of the appearance count DB 23, and the value criterion and rank criterion of the evaluation criterion DB have already been generated.

  In step S <b> 1, the control unit 300 receives an input of an evaluation target article. Here, a plurality of articles to be evaluated are received, and articles to be presented to the user terminal 40 are selected in subsequent processing.

  In step S2, the control unit 300 classifies one of the evaluation target articles received in step S1 as one of the normalized patterns stored in the pattern DB 22.

  In step S3, the control unit 300 extracts a corresponding phrase from the evaluation target article for the evaluation target items included in the normalization pattern classified in step S2.

  In step S4, the control unit 300 refers to statistical data stored in the appearance count DB 23 for the phrase extracted in step S3, and calculates a deviation value of the appearance count. Further, the control unit 300 refers to the value standard stored in the evaluation standard DB 24 and calculates the importance of the word / phrase.

  In step S5, the control unit 300 further weights the importance of the phrase calculated in step S4 based on the value criterion stored in the evaluation criterion DB 24, and evaluates it as an overall score of the entire evaluation target article. Calculate the value.

  In step S6, the control unit 300 determines whether there is a next evaluation target article, that is, whether there is an evaluation target article that has not been calculated in the evaluation target article received in step S1. If this determination is YES, the process proceeds to step S7, and if the determination is NO, the process returns to step S2 to continue the process for the next evaluation target article.

  In step S7, the control unit 300 selects a predetermined number of articles whose evaluation values calculated in step S5 are higher than the evaluation target articles received in step S1. Note that articles whose evaluation value is less than the predetermined value may be excluded. In this case, the number of articles to be selected is less than a predetermined number, but it is possible to suppress the selection of articles with low value.

  In step S8, the control unit 300 ranks the articles selected in step S7 in descending order of evaluation value. At this time, if there are articles having the same evaluation value, the rank of articles including more important words is raised based on the rank criterion stored in the evaluation criterion DB 24.

  In step S9, the control unit 300 selects a word / phrase to be used for title data for the article selected in step S7. Specifically, a predetermined amount of phrases are selected in descending order of importance of the weighted phrases.

  In step S10, the control unit 300 connects the phrases selected in step S9 to generate article title data. The control unit 300 provides the article data together with the title data to the Web server 30. Thereby, the high-value article is displayed together with the title on the user terminal 40 that has accessed the Web page managed by the Web server 30.

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

DESCRIPTION OF SYMBOLS 10 Evaluation server 11 Article pattern production | generation part 12 Article reception part 13 Article pattern classification | category part 14 Phrase extraction part 15 Value evaluation part 16 Attention phrase selection part 17 Title generation part 18 Value reference | standard reception part 19 Ranking reference | standard reception part 20 Content server 21 Article DB
22 Pattern DB
23 Appearance count DB
24 Evaluation DB
30 Web server 40 User terminal 300 Control unit 410 Storage device

Claims (10)

An apparatus for evaluating the value of an article,
Pattern storage means for storing one or more normalized patterns that include a combination of items included in past articles;
Classification means for classifying the article to be evaluated and the past article into any of the normalized patterns stored by the pattern storage means;
Extraction means for extracting words in the evaluation target article and the past article corresponding to one or more items included in the normalized pattern classified by the classification means;
For each of the phrases extracted by the extraction means, an appearance count storage means for storing statistical data associated with the appearance count in the evaluation target article and the past article;
In the statistical data stored by the appearance count storage means, a calculation means for calculating an evaluation value indicating the value of the evaluation target article based on a deviation of the appearance count of the words included in the evaluation target article. apparatus. A value criterion accepting unit that accepts an input of a value criterion that defines the importance of the phrase with respect to the deviation of the number of appearances;
The apparatus according to claim 1, wherein the calculation unit calculates the evaluation value by combining the importance of words for each item included in the evaluation target article. Selection means for comparing the importance of words for each item included in the article to be evaluated, and for selecting words with relatively high importance;
The apparatus according to claim 2, further comprising: generation means for generating title data composed of words selected by the selection means. The selecting means selects a predetermined amount of words / phrases set in advance,
The apparatus according to claim 3, wherein the generation unit generates the title data by arranging words selected by the selection unit in order of the importance.   The apparatus according to any one of claims 2 to 4, wherein the value criterion accepting unit accepts an input of weighting of the importance level that is different for each item as the value criterion.   6. The apparatus according to claim 2, wherein the value criterion accepting unit accepts, as the value criterion, an input of the weighting of the importance level that is different for each acquisition source of the evaluation target article. .   The apparatus according to any one of claims 2 to 6, wherein the value criterion accepting unit accepts an input of the weight of the importance level that is different for each normalization pattern as the value criterion. A rank reference receiving means for receiving an input of a rank reference that defines the priority of a plurality of words corresponding to the same item;
The apparatus according to claim 1, further comprising: a determination unit that determines the ranks of a plurality of articles having the same evaluation value based on the rank criterion. A method by which a computer evaluates the value of an article,
A pattern storage step of storing one or more normalized patterns that are configured to include combinations of items included in past articles;
A classification step of classifying the article to be evaluated and the past article into any of the normalized patterns stored in the pattern storage step;
An extraction step of extracting words in the evaluation target article and the past article corresponding to one or more items included in the normalized pattern classified by the classification step;
For each of the phrases extracted in the extraction step, an appearance count storage step for storing statistical data associated with the appearance count in the evaluation target article and the past article;
In the statistical data stored in the appearance count storage step, a calculation step of calculating an evaluation value indicating the value of the evaluation target article based on a deviation of the appearance count of the words included in the evaluation target article is included. Method.   The program which makes a computer perform the method of Claim 9.

Images