JP5168961B2 - Latest reputation information notification program, recording medium, apparatus and method - Google Patents

Description

The present invention relates to a latest reputation information notification program, a recording medium, an apparatus, and a method for extracting and notifying the latest reputation information from a web document on the Internet, and more particularly to the latest articles related to user registration keywords on the web. The present invention relates to a latest reputation information notification program, a recording medium, a device, and a method for retrieving an overview from a combination of an evaluation expression and a notification.

  In recent years, reputation information on their products and services is important information for companies. Above all, negative opinions are very important. If you have a bad reputation in part, if you can know it quickly and respond quickly, you can prevent it from spreading into a big problem. it can.

  Conventionally, there is a reputation information processing method that collects text on the web such as a blog, extracts reputation information written by general users, and analyzes it (Non-Patent Document 1). For example, in the case of reputation analysis for “Company F”, a text in which an opinion about F Company is written is searched to determine whether the sentence is positive or negative.

The main purpose of this reputation information processing method is to collect a large number of opinions from the web, perform statistical analysis, and use it for marketing. For example, regarding the personal computer manufactured by Company F, what percentage of the opinion is “easy to use” and what percentage is “difficult”, the graph of changes in the number of reputations over time, and the comparison of opinions with multiple products It is displayed in the form of an anchor map that looks at the distribution.
JP 11-219375 A JP 2004-157841 A “Nifty marketing service”, [February 6, 2007 search], Internet (URL: http://www.nifty.com/buzz/)

  However, in such a conventional reputation information processing method, a large number of opinions are collected as reputation information from the web to be used for marketing, and statistical analysis is performed. The problem is that it is not suitable for the usage that wants to grasp each item as soon as it comes out, for example, to check the most important negative articles in the latest negative reputation information first in the morning. There is.

  An object of the present invention is to provide a latest reputation information notification program, a recording medium, a method, and an apparatus that can extract reputation information from articles published on the web and efficiently grasp the latest reputation information.

Another object of the present invention is to provide a latest reputation information notification program, recording medium, apparatus, and method that can efficiently grasp malicious articles such as slander and abuse on the web.

(program)
The present invention provides a latest reputation information notification program. The latest reputation information notification program of the present invention is a computer,
A keyword registration section that registers objects that you want to know about as a keyword,
A search unit that periodically searches documents existing on the Internet for keywords,
Analyzing the document searched by the search unit to extract a reputation pair that combines the object registered as a keyword and a reputation word, calculate a score indicating the importance of the reputation pair, and store the document together with the reputation word and the score in the database Reputation information extractor that registers summary information for
Reputation information display section that displays summary information of documents in the database ranked by score,
It is made to function as.

Here, the search unit
The latest information search part that searches the latest information by visiting the information sources on the Internet,
A polling unit that periodically operates the latest information search unit to search for documents including keywords,
Is provided.

The reputation information extraction unit
A reputation pair generation unit that extracts one or more reputation words from the retrieved documents and generates a reputation pair that is combined with an object registered as a keyword;
A score calculator that calculates a score based on one or more reputation pairs included in the document;
Database registration for registering summary information including an information source address (URL) of a document, an identifier indicating whether the reputation is positive reputation (positive) or negative reputation (negative), a title, and an information source name together with the reputation word and score in the database And
Is provided.

  The score calculation unit obtains a syntax score, a word score, an appearance frequency score, and an information source score for each reputation pair in the document for each document, and calculates a total score used for ranking based on the score.

The syntax score is a value indicating the certainty of the reputation pair calculated based on the features in the document of the object and reputation word constituting the reputation pair,
The word score is a value indicating the strength of expression of reputation words set in advance,
The appearance frequency score is a value indicating the number of reputation words in the document,
The information source score is a value representing the reliability of a document as an information source (the magnitude of influence when a bad reputation comes out).

  The present invention further provides a program for causing a computer to function as a browsing history reflection unit that updates the information source score using a browsing history of documents on a network at regular intervals.

  In addition, the browsing history reflection unit extracts, from the browsing history, information source document addresses that have not been registered in the search unit's circulation destination so far, and adds them to a new circulation destination.

  The reputation display unit ranks the summary information of the documents in the database according to the score, displays the newly arrived information, and manages unread and read.

(recoding media)
The present invention provides a computer-readable recording medium storing a latest reputation information notification program. The recording medium of the present invention includes a computer,
A keyword registration section that registers objects that you want to know about as a keyword,
Analyzing the document searched by the search unit to extract a reputation pair that combines the object registered as a keyword and a reputation word, calculate a score indicating the importance of the reputation pair, and store the document together with the reputation word and the score in the database Reputation information extractor that registers summary information for
Reputation information display section that displays summary information of documents in the database ranked by score,
It is made to function as.

(apparatus)
The present invention provides a latest reputation information notification device. The latest reputation information notification device of the present invention is
A keyword registration unit that registers the target object of the reputation as a keyword,
A search unit that periodically searches documents on the Internet for keywords,
Analyzing the document searched by the search unit to extract a reputation pair that combines the object registered as a keyword and a reputation word, calculate a score indicating the importance of the reputation pair, and store the document together with the reputation word and the score in the database A reputation information extractor for registering summary information,
Reputation information display section that displays summary information of database documents ranked by score,
It is provided with.

(Method)
The present invention provides a latest reputation information notification method. The latest reputation information notification method of the present invention includes:
A keyword registration step for registering an object whose reputation is desired as a keyword,
A search step for periodically searching documents existing on the Internet for keywords;
Analyzing the document retrieved in the search step, extract a reputation pair that combines the object registered as a keyword and a reputation word, calculate a score indicating the importance of the reputation pair, and document the reputation word and score in the database Reputation information extraction step to register the summary information of
Reputation information display step for ranking and displaying summary information of database documents by score,
It is provided with.

(program)
According to another aspect of the present invention, there is provided a latest reputation information notification program for notifying a latest document including a cautionary word of interest. The latest reputation information notification program of the present invention is a computer,
A keyword registration section that registers reputable words as keywords,
A search unit that searches documents on the Internet for keywords,
Analyzes the documents searched by the search unit, extracts the reputation pairs that combine the cautionary reputation words registered as keywords and the object, calculates the score indicating the importance of the reputation pair, and calculates the cautionary reputation words in the database. And a reputation information extraction unit for registering document summary information together with the score,
Reputation information display section that displays summary information of documents in the database ranked by score,
It is made to function as.

  The latest reputation information notification program according to the present invention further extracts a co-occurrence word related to a reputable attention word registered as a keyword from a searched document and registers it in a database, and the reputation information display unit shares it. It is made to function as a related keyword classification | category part which displays a word as a related keyword with summary information.

(recoding media)
According to another aspect of the present invention, there is provided a computer-readable recording medium storing a latest reputation information notification program for notifying a latest document including a reputable word of interest. The recording medium of the present invention includes a computer,
A keyword registration section that registers reputable words as keywords,
A search unit that searches documents on the Internet for keywords,
Analyzes the documents searched by the search unit, extracts the reputation pairs that combine the reputable word of interest registered as keywords and the object, calculates the score indicating the importance of the reputation pair, and writes the score to the database. Reputation information extraction unit that registers summary information of documents together with words and scores,
Reputation information display section that displays summary information of documents in the database ranked by score,
It is made to function as.

(apparatus)
According to another aspect of the present invention, there is provided a latest reputation information notifying device for notifying a latest document including a reputable word of interest. The latest reputation information notification device of the present invention is
A keyword registration section that registers reputable words as keywords,
A search unit for searching documents existing on the Internet for keywords;
Analyzes the documents searched by the search unit, extracts the reputation pairs that combine the cautionary reputation words registered as keywords and the object, calculates the score indicating the importance of the reputation pair, and calculates the cautionary reputation words in the database. A reputation information extraction unit for registering document summary information together with a score;
Reputation information display section that displays summary information of database documents ranked by score,
It is provided with.

(Method)
According to another aspect of the present invention, there is provided a latest reputation information notification method for notifying a latest document including a cautionary word of interest. The latest reputation information notification method of the present invention includes:
A keyword registration step for registering reputable words as keywords,
A search step for searching documents on the Internet for keywords;
Analyzes the documents searched by the search unit, extracts the reputation pairs that combine the cautionary reputation words registered as keywords and the object, calculates the score indicating the importance of the reputation pair, and calculates the cautionary reputation words in the database. And a reputation information extraction step for registering document summary information together with the score;
Reputation information display step for ranking and displaying summary information of database documents by score,
It is provided with.

  According to the present invention, when an article about a keyword (object) registered by a user appears in a blog, news, etc. on the Internet, in addition to automatically searching and notifying the user as a new article, By extracting reputation pairs that combine objects and reputation words, ranking by the score indicating the importance of reputation pairs, and displaying the distinction between read and unread and summary of articles, users can find out about the keywords they are interested in It is possible to grasp the latest reputation information in a very efficient manner, that is, in a short time and in a large amount without fail.

  For this reason, for example, if there is a bad reputation for some of your products or services, you can know it quickly and respond quickly, and you can prevent it from spreading to major problems. .

  In another form of the present invention, if an article containing a malicious reputation word registered as a keyword that causes slander or human rights violations appears on a blog or news on the Internet, the document is searched by keyword and the target object is searched. Reputation pairs that combine reputation words are extracted and summaries of articles that are likely to be unread slander slander or human rights violations ranked by scores indicating the importance of reputation pairs are displayed, and the latest problematic reputation information is displayed Therefore, it is possible to grasp all of them in a large amount in a short time.

Furthermore, since the browsing history is fed back (reflected) to the search and the score, the search performance of reputation information can be improved as it is used.
(Comment: Claim 6 is obtained by adding a history reflection part to Claim 1 and is at the same level as Claim 1 and is not an embodiment claim that limits Claim 1, so the effect is described. )

  FIG. 1 is a block diagram of a functional configuration showing an embodiment of the latest reputation information notification device according to the present invention. In FIG. 1, the latest reputation information notification device 10 of the present embodiment is a function realized by executing a program in the hardware environment of a computer as shown in FIG. 2, and includes a keyword registration unit 22, a keyword list 24, and latest information. Search unit 26, polling unit 28, reputation information extraction unit 30, reputation word dictionary 32, search document reputation pair database 34, search document reputation word database 36, information search score counter 38, reputation information display unit 40, reputation notification database 42 The browsing history database 44 and the browsing history reflection unit 46 are provided.

  The latest reputation information notification device 10 of the present embodiment has a function of automatically collecting and notifying the reputation of the company or product desired by the user on the Internet as the latest reputation information.

  For this reason, the latest reputation information notification device 10 is connected to the Internet 12 and collects the latest articles as documents existing on sites such as the blog site 14, the news site 16, the word-of-mouth site 18, and the bulletin board site 20 on the Internet 12. And notify the reputation information.

  The keyword registration unit 22 provided in the latest reputation information notification apparatus 10 registers, as a keyword, a name that identifies a target for which a user using this embodiment wants to know the reputation. For example, when it is desired to know the reputation of the company or the reputation of the company's product, the company name and product name are registered as keywords. The keywords registered by the keyword registration unit 22 are stored in the keyword list 24.

  The latest information search unit 26 and the polling unit 28 constitute a search unit. The latest information search unit 26 is a search engine for the latest article. For example, it is a full-text search type search engine known as a crawler. The content of a web page is stored in a database on the search engine side, and a search request is made. When there is, it searches the datahase and returns the result. The crawler program functions as a search robot and, while following the patrol destination, discovers web pages that are not registered in the database and updated web pages, collects the contents, and reflects the results in the database.

  The polling unit 28 periodically searches and acquires the latest document including the keyword from the circulation destination site registered in advance by the operation of the latest information search unit 26 using the keyword registered in the keyword list 24. .

  The reputation information extraction unit 30 analyzes the document searched by the polling unit 28, extracts a reputation pair that combines a target such as a company name or a product name and a reputation word obtained by searching the document, and each document Each time, a score indicating the importance of the reputation pair is calculated, and the summary information of the document is registered in the database together with the reputation word and the score.

  In the present embodiment, a reputation word dictionary 32, a search document reputation pair database 34, a search document reputation word database 30, and a reputation notification database 42 are provided as databases for the reputation information extraction unit 30.

  The reputation word dictionary 32 is used when a reputation word is extracted by analysis of a document in the reputation information extraction unit 30. The search document reputation pair database 34 stores reputation pairs in which the object extracted by the reputation information extraction unit 34 and reputation words are combined. The search document reputation word database 36 stores the frequency of reputation words of the search document extracted by the reputation information extraction unit 30.

  In the reputation notification database 42, summary information of the document including the reputation word and score extracted by the reputation information extraction unit 30 is registered. The summary information of this document includes the URL that is the information source address of the document together with the reputation word and score, the identifier P / N indicating whether the reputation is positive (positive reputation) or negative (negative reputation), the title of the document, and the information source For example, information including a news source name is registered.

  The reputation information display unit 40 ranks the summary information of documents registered in the reputation notification database 42 based on the score and displays the ranking information on the display. By displaying the summary information of search documents ranked by the score indicating the importance of this reputation against the user along with the distinction between unread and read, for example, reputation information about the company and its products, especially bad reputation information Even in the event of a problem, it is possible to quickly know the bad reputation information and take a quick and appropriate response to prevent it from spreading to a major problem. You can prevent you from seeing an article you have already read many times, and it takes less time to check the information.

  For example, a person in charge of a company browses the latest reputation information of the company and product according to the present embodiment first after coming to the office in the morning, and this is quickly performed when bad reputation information is output. To understand and take appropriate action.

  FIG. 2 is a block diagram showing a hardware environment of a computer in which the latest reputation information notification program according to the present embodiment is executed. In FIG. 2, a RAM 52, a ROM 54, a hard disk drive 56, a keyboard 60, a mouse 62, a device interface 58 for connecting a display 64, and a network adapter 66 are connected to the bus 50 of the CPU 48.

  The hard disk drive 56 is installed with a program that realizes the function of the latest reputation information notification device 10 of FIG. When the computer is started, the OS is expanded from the hard disk drive 56 to the RAM 52 by the boot process by the BIOS of the ROM 54, and then the latest reputation information notification program of the present embodiment is read from the hard disk drive 56 to the RAM 52 by the execution of the OS, and the CPU 48 It is executed by.

  FIG. 3 is an explanatory diagram showing a user screen on which the latest reputation information notification is displayed in the present embodiment. In FIG. 3, the display area of the keyword list 72, the article summary list 75, and the browser 78 is provided on the user screen 70.

  The keyword list 72 displays a keyword list registered by the user, such as “Company F”, “Company B”, and “Notebook PC”. Here, folders are divided into a positive reputation (positive reputation) and a negative reputation (negative reputation).

  The numbers in [] to the right of the folder and keyword indicate the number of retrieved documents. In this example, the new arrival display 74 “Company F [2]” indicates that there are two latest negative reputations for Company F.

  When one element of the keyword list 72 is selected (in the figure, the selected keyword is surrounded by a frame), the article summary list 75 displays the title, news source, and update of the article extracted as reputation information about the selected keyword. The date / time, reputation word, and importance items are displayed. As for the contents of the title, as shown by the envelope mark on the left side, whether it has been read or not is displayed. In this example, two of the article outlines 76-1 and 76-2 are unread, and the item fields are displayed in different colors according to the importance of the article.

  For example, an article summary 76-2 with a low importance level of “No 0.5” is displayed in a lighter color, and an article outline 76-1 with a high importance level “No 0.7” is displayed with a darker color. I try to make it stand out.

  In the area of the browser 78 below the user screen 70, for example, the web page of the article with the highest importance in the article summary list 75 is displayed as a default. The article displayed on the browser 78 is underlined, for example, with markings 80 in the portion having reputation information, so that the portion to be checked can be quickly seen.

  FIG. 4 is an explanatory diagram showing another example of the user screen informed of the latest reputation information of the present embodiment. On the user screen 82 in FIG. 4, article summaries 76-1 and 76-2 indicating the latest reputation information obtained by the processing of this embodiment are displayed as a series of descriptive documents without particular item names. ing.

  For example, the article summary 76-1 is the same as the user screen 70 in FIG. 3, but on the user screen 82 in FIG. 4, the title “I bought a personal computer ...” is a news source “ "Fujiko Kawasaki's blog", update date and time "9.20 AM 11:00" and importance level "Negation degree 0.7" are described in a series of documents, and further displayed by browser 78 in FIG. I picked up the part of the article that includes the reputation word and bought "Company F", but the support center is the worst! Is described.

  For either the user screen 70 of FIG. 3 or the user screen 82 of FIG. 4, the article summary is displayed in order of importance for the latest reputation information obtained in this embodiment, and the user has registered keywords. The latest reputation information about the company name and product name can be checked appropriately.

  FIG. 5 is a flowchart showing the latest reputation information notification process according to the embodiment of FIG. 1 and will be described below with reference to FIG.

  In FIG. 5, first, it is checked in step S1 whether or not there is a keyword input by the user. If there is a keyword input, the process proceeds to step S2, and an object that the user who has input as a keyword wants to know about the reputation is determined for each user. Register in list 24.

  FIG. 6 shows the contents of the keyword list 24 of FIG. The keyword list 24 includes a list ID, a user ID, and a keyword. In the keyword list 24, information for identifying an object such as a company name or a product name that the user wants to know about the reputation is registered.

  For example, in the list ID = 001, “kawafuji” is registered as the user ID, and for this user, the company name “Company F” and the product name “PQ server” are used as keywords for specifying the object whose reputation is to be known. Registered.

  Referring again to FIG. 5, when the user keyword registration is completed in step S <b> 2, the process proceeds to step S <b> 3, where it is determined whether or not the polling unit 28 has reached a predetermined polling cycle. Then, the process proceeds to step S4, and for the keyword obtained from the keyword list 24, the latest information search unit 26 is used to search for a document containing the keyword for the currently registered site on the Internet 12 as a target.

  In step S5, reputation information extraction processing is executed for the retrieved document, the retrieved document is analyzed, and a reputation pair that combines the object registered as a keyword and a reputation word is extracted. The score indicating the importance of the reputation pair is calculated, and the summary information of the document is registered together with the reputation word and the score in the reputation notification database 42 in step S6.

  Subsequently, if there is a user browsing request in step S7, the reputation information display unit 40 reads the reputation information corresponding to the user from the reputation notification database 42 in step S8, and compares it with the browsing history, for example, the user screen 70 in FIG. Or it displays as shown in the user screen 82 of FIG.

  Subsequently, in step S9, for example, it is checked whether or not it is a batch processing cycle performed once a day in the night. When the batch processing cycle is determined, a browsing history reflection process is executed in step S10. In the browsing history reflection process, the upper address of the information source, that is, the upper URL, is extracted from the browsing history of the document on the network obtained within a certain period, for example, one month before, and stored in the information source score counter database 38. The value of the information source score in the information source score counter is updated, and further, the information source document address not registered in the circulation destination of the latest information search unit 26 is extracted and added to the new circulation destination.

  Such processes of steps S1 to S10 are repeated until a stop instruction is issued in step S11.

  FIG. 7 is a flowchart showing details of reputation information extraction processing in step S5 of FIG. Reputation information extraction processing in the present embodiment analyzes each document obtained as a search result by keyword, extracts a reputation pair that combines a target object and a reputation word, and obtains a score indicating importance for the reputation pair After that, it is stored in the reputation notification database, and the result is ranked and sorted by score.

  Referring to FIG. 7, the reputation information extraction process will be described in more detail. First, a retrieval document is extracted in step S1, a document is analyzed in step S2, and a reputation pair composed of a combination of an object and a reputation word is generated from the analysis result. To create a database. This database generates two, the search document reputation versus database 34 and the search document reputation word database 36 of FIG.

  Subsequently, in step S3, it is checked whether or not there is a reputation word related to the object that is the designated keyword in the extracted reputation pair. If there is a corresponding reputation pair, the process proceeds to step S4, and a score giving process for giving a score indicating the importance of the reputation pair is executed.

  In step S5, the summary information of the document including the reputation word and score extracted in the reputation notification database 42 is stored. Subsequently, in step S6, it is checked whether or not a search document by polling remains. If it remains, the process returns to step S1 again, and the same processing is repeated for the next document. If no search document remains, the process proceeds to step S7, the documents registered in the reputation notification database 42 are rearranged in the score order, and the process returns to the main routine of FIG.

  FIG. 8 is a flowchart showing details of the reputation pair extraction process in step S2 of FIG. In FIG. 8, the reputation pair extraction process performs morphological analysis on the document searched in step S1. By this morphological analysis, the search document is divided into words.

  Note that the morphological analysis of the search document is used in the case of a language in which words are not separated by a space, such as Japanese, Chinese, and Hangul, while the words such as English and German are used. This is not necessary for documents whose language is separated by a space.

  Subsequently, in step S2, an object such as a company name or a product name is extracted from the search document after morphological analysis. To extract the target object, a process of extracting a specific expression from the search document as a target object and a process of extracting a noun phrase as the target object are performed. Further, a general noun is extracted using a dictionary (not shown). To do.

  Subsequently, in step S3, reputation words for the object are extracted from the search document using the reputation word dictionary 32. In the reputation word dictionary 32, negative reputation words such as “worst”, “slow”, “minimum”, positive reputation words such as “like”, “pretty”, “great”, and none belong to any of them. Other reputation words are registered in advance.

  An evaluation word is extracted by matching the reputation word registered in the reputation word dictionary 34 with the phrase of the search document. Furthermore, in the present embodiment, for the reputation words registered in the reputation word dictionary 32, a word score indicating the degree of reputation and an identifier indicating positive or negative are used. Has registered P.

  Subsequently, after extracting a reputation pair combining the object extracted from the search document and the reputation word in step S4, the syntax score of the reputation pair is determined in step S5. The syntax score of reputation pair is a value indicating the probability, and a feature amount is extracted from the extracted reputation pair, and this is used as a syntax score.

Here, the feature quantity for detecting the syntactic score extracted for reputation pairs is as follows: (1) Distance between reputation words and objects (2) Kinds of reputation words and objects (3) Reputation words and surroundings of objects .., Fn are obtained, weights w1, w2,... Wn are set for the respective feature amounts, and the syntax score C1 is calculated by the following equation, for example.
S1 = w1f1 + w2f2 + ... wnfn
As the weights w1 to wn in this calculation formula, values obtained in advance for the search document with the model are used.

  If the reputation pair syntax score C1 can be determined in step S5, the process proceeds to step S6, and necessary information is stored in the search document reputation pair database 34 and the search document reputation word database 36 of FIG.

  FIG. 9 is an explanatory diagram of the search document reputation pair database 34 of FIG. 1 in which necessary information is stored by executing the reputation pair extraction process of FIG. In FIG. 9, the search document reputation pair database 34 includes an ID, a URL indicating an information source address, a reputation pair, a positive / negative, and a syntax score.

  For example, for ID = 001, “http://a-blog.jp/0661030.html” is registered as the URL, and “F company, worst” is registered as the reputation pair extracted for the document acquired from this URL. ing. This reputation pair is “Company F” whose target is a company name, and the reputation word extracted from a document searched using this as a keyword is “worst”.

  Subsequently, “N” indicating negative as positive / negative is registered, and finally, a syntax score “0.3” determined for reputation vs. “Company F, worst” is stored.

  Here, the syntax score indicates the certainty of reputation pair. For example, when a document “F company is worst” is processed, “F company, worst” is obtained as reputation pair. In addition, even if a document “Buy a personal computer from the website of Company F. This is the worst thing to hear” is processed, “Company F, worst” is also extracted as a reputational pair.

  For such two reputation pairs, the syntactic score is given a high syntactic score because the distance between the object “Company F” and the reputation word “worst” is short for the former. On the other hand, since the distance between the object “Company F” and the reputation word “worst” is far, the latter is given a low score.

  FIG. 10 is an explanatory diagram showing the contents of the search document reputation word database 36 of FIG. 1 generated by the reputation pair extraction process of FIG. The search document reputation word database 36 of FIG. 10 stores the reputation words obtained by the reputation pair extraction process of the search document together with the URL of the information source.

  That is, the search document reputation word database 36 includes ID, URL, reputation word, word score, positive / negative, and appearance frequency. For example, when a reputation pair as shown in the search document reputation pair database 34 of FIG. 9 is extracted by the reputation pair extraction process, the same URL is registered, the reputation word “worst” is registered, and the reputation word dictionary 32 is further registered. The word score “0.9” obtained from the above is registered, “N” indicating negative is registered as positive / negative, and “1” is registered as the appearance frequency.

  Here, the word score in the search document reputation word database 36 is used as the word score C2, and the appearance frequency is used as the frequency score C3 in the calculation of the total score C described later.

  FIG. 11 is a flowchart showing details of the score assignment process in step S4 of FIG. In FIG. 11, in the score assignment process, first, in step S1, the URL item specified by the URL of the target document is extracted from the search document reputation pair database 34 shown in FIG.

  Subsequently, the process proceeds to step S2, and it is determined whether or not there is a reputation pair having a syntax score C1 equal to or greater than a predetermined threshold among the items extracted from the search document reputation pair database 34.

  If there is a reputation pair whose syntax score C1 is greater than or equal to the threshold value, the process proceeds to step S3, and a word score C2 and an appearance frequency score C3 for the reputation word included in the reputation pair are extracted from the search document reputation word database 36 shown in FIG. To do. In step S4, the information source score C4 of the target URL is calculated. Finally, the total score C is calculated by the following equation in step S5.

この（１）式において、右辺第１項は総スコアＣを０〜１に正規化する関数である。

In this equation (1), the first term on the right side is a function that normalizes the total score C to 0-1.

  FIG. 12 is a flowchart showing details of the information source score calculation process in step S4 of FIG. In this information source score calculation process, an information source score counter table 38-1 shown in FIG. 13 and an information source score table 38-2 shown in FIG. 14 are used. The information source score counter table 38-1 and the information source score table 38-2 in FIGS. 13 and 14 are arranged in the information source score counter 38 in FIG.

  The information source score counter table 38-1 in FIG. 13 includes an ID, a higher URL, and a browsing frequency. In principle, the upper URL of the information source score counter table 38-1 is extracted and registered up to the first level of “http: // host name / xxx /”, but if not, “http: // / Host name / ". The browsing frequency is the number of browsing times collected in units of users from the present time to, for example, one month ago.

  The information source score table 38-2 of FIG. 14 stores the score calculated by the information source score calculation process of FIG.

  The information source score calculation process in FIG. 12 will be described as follows. First, in step S1, the upper URL of the target URL is extracted by the same method as the information source score counter table 38-1 in FIG. 13, and it is determined whether or not the URL has already appeared in the information source score counter table 38-1. If it has appeared, the process proceeds to step S2, and the browsing frequency A1 is extracted.

In step S3, the highest browsing frequency A2 is extracted from the information source score counter table 38-1. Subsequently, by dividing the browsing frequency A1 by the maximum browsing frequency A2 in step S4,
S4 = (A1 / A2)
As a result, a normalized information source score C4 is calculated.

  On the other hand, if the upper URL of the target URL does not exist in the information source score counter table 38-1 in step S1, the process proceeds to step S5, and the total number n of frequencies in the information source score counter table 38-1 is obtained. 1 / n) is assigned as the minimum information source score C4.

  FIG. 14 shows an information source score initial value table 38-2 storing the information source score C4 calculated by the processing of FIG. 12 based on the information source score counter table 38-1 of FIG. In this case, for the target URL that matches the upper URL of ID = 001, the score C4 = 1.0 is obtained with A1 = 128 and A2 = 128.

  For ID = 002, A1 = 59 and A2 = 128, and the information source score C4 = 0.5 (rounded to the second decimal place) is determined. Furthermore, since ID = 003 is a target URL that does not have an upper URL that matches the information source score counter table 38-1, the total number of frequencies n = 205 is obtained and 1 / n = 0.004878. It is set to 0.0 by rounding off the second decimal place.

  In addition, for a specific site, a rule for assigning a predetermined score may be set separately without calculating by this method. Another method is to record the appearance frequency of the article and calculate the score based on the click rate (viewing frequency / occurrence frequency) of the article.

  FIG. 15 is an explanatory diagram showing the contents of the reputation notification database generated by the process of step S6 of FIG. The reputation notification database 42 of FIG. 5 registers ID, URL, positive / negative, title, news source, update date, reputation word, and total score.

  Among these, URL, positive / negative, reputation word, and total score are information obtained by the reputation information extraction process in step S5 of FIG. For document summary information consisting of other titles, news sources, and update dates and times, if the target document has an RDF (Resource Description Framework) corresponding to an XML-format catalog, subtitles, article titles, and updates from this RDF The date and time are extracted and registered in the reputation notification database 42.

  If there is no RDF in the target document, the title and update date of the article are extracted using the HTML title tag and the file update date and time and registered in the reputation notification database 42. If the site title is unknown, the article title and site title may be extracted by inductive reasoning (heuristic rules).

  When there is a user browsing request in step S7 in FIG. 5, the corresponding data is read from the reputation notification database 42 shown in FIG. 15, and the user screen 70 in FIG. 3 or the user screen 82 in FIG. 4 is displayed. The latest reputation information about the company name and product name, which are objects registered by the user as keywords, can be browsed.

  The information in this database may be provided in a form that can be read by a general RSS reader in the form of RSS (RDF Site Summary) for distributing site update information and the like.

  FIG. 16 is a flowchart showing details of the browsing history reflection process in step S9 of FIG. In the latest reputation information notification device 10 according to the present embodiment, while the user repeatedly browses the website, the ranking of the notified importance level is higher, but this site does not have to be viewed. Judgment may become apparent as you use it.

  Therefore, in this embodiment, browsing history reflection processing is performed in which the user's browsing history is fed back to re-evaluate the information source score C4 that affects the calculation of the total score C. Furthermore, simultaneously with the re-evaluation of the information source score based on the browsing history, the latest information search unit 26 extracts a search destination not included in the circulation destination from the browsing history and newly adds it.

  In the browsing history reflection process of FIG. 16 for this purpose, first, the already read URL is extracted from the browsing history database in which the browsing history within a certain period is registered in step S1, and the upper URL is extracted from the URL in step S2. .

  FIG. 17 shows the contents of the browsing history database 44 referred to in step S1. The browsing history database 44 includes an ID, a user ID, a read URL, and a date.

  Referring to FIG. 16 again, for the upper URLs in the browsing history database extracted in step S2, the same upper URLs of the same user are combined into one in step S3. Subsequently, in step S4, the browsing frequency is totaled for each upper URL, and updated by replacing the browsing frequency with the current browsing frequency of the information source score counter table 38-1 shown in FIG. 13, for example.

  By updating the information source score counter 38 based on such browsing history, for example, the information source score C4 used for calculating the total score C is a high value before the update for the site that has been ranked high in the ranking so far. As a result, the total score C is lowered and the ranking can be lowered.

  Subsequently, in step S5, the URL of the updated information source score counter is output to the latest information search unit 26 operating as the search engine of FIG. That is, in the latest information search unit 26, if there is a URL notified from the browsing history reflection unit 46 that is not included in the currently crawling destination URL, this is added to the crawling destination. To do.

  The browsing history reflection processing and the crawling circulation destination addition processing in the latest information search unit 26 based on the URL notified from the browsing history reflection unit 46 are executed as a batch process once a day, for example, at night. .

  In the update of the information source score counter table in step S4 of FIG. 16, for URLs that have appeared in the previous update but not in the current update, S4 = 0 is stored as the information source score.

  FIG. 18 is a flowchart showing another embodiment of the score assignment process in step S4 of FIG. In the document on the web targeted by the present embodiment, for example, a large number of comments may be concentrated on a certain blog article (in the above-mentioned state). It is thought that many readers sympathize with or repel the content of the article, and the importance as reputation information is high.

  In the score assigning process of FIG. 18, scoring is performed with emphasis on the above-mentioned flames. 18, steps S1 to S4 are the same as steps S1 to S4 in FIG. 11. When the syntax score C1 is greater than or equal to the threshold value for the reputation pair obtained from the URL item of the target document, the word score C2 and the appearance frequency score C3. Further, an information source score C4 is obtained.

  In addition to this, in the above-described search for flames, in step S5, among the search results by the latest information search unit 26, a document with N or more predetermined comments attached to the search document, for example, N = 1000 or more is attached. Is extracted as a comment count score C5. In step S6, the total score C is calculated by the following equation including the newly extracted comment number score C5.

このように炎上記事検索によるコメント数スコアＣ５を加えた総スコアＣを計算することで、コメントが集中している記事を重要な評判情報として表示、閲覧することができる。

Thus, by calculating the total score C including the comment number score C5 obtained by the above-mentioned flame search, it is possible to display and browse articles with concentrated comments as important reputation information.

  FIG. 19 is a block diagram of a functional configuration showing another embodiment of the latest reputation information notifying apparatus according to the present invention. In this embodiment, the latest article of the attention reputation word is searched. And

  This is because there may be articles such as slander and human rights violations in articles on the Internet, and the latest information as well as reputation information about such slander and human rights violations articles. This embodiment needs to be checked and corresponds to this.

  19, the latest reputation information notification device 100 includes a keyword registration unit 22, a keyword list 240, a latest information search unit 26, a polling unit 28, a reputation information extraction unit 30, a reputation word dictionary 32, a search document reputation versus database 34, A search document reputation word database 36, an information source score counter 38, a reputation information display unit 40, a reputation notification database 420, a browsing history database 44, and a browsing history reflection unit 46 are provided.

  Among these, as a unique configuration in the embodiment of FIG. 19, a related keyword classification unit 84 and a co-occurrence word database 86 are newly provided as compared with the example of FIG. 1. Further, in the keyword list 240 and the reputation notification database 420, the keyword to be registered is a reputable caution word, and the reputation notification information is an article such as slander or human rights violations. have.

  FIG. 20 is an explanatory view showing a user screen for notifying the latest reputation information for the slanderous articles and human rights violation articles displayed by the embodiment of FIG.

  In FIG. 20, the user screen 88 is divided into a related keyword list 90, an article summary list 92, and a browser 96. The article summary list 92 includes article summaries 94-1 and 94-2 created by reputation information extraction processing by searching from the Internet using “criminals”, “leave”, etc., which are important keywords registered by the user. 94-3 is ranked and displayed by the importance which is the calculated score.

  The article summaries 94-1 to 94-3 display titles, news sources, update dates and times, caution words that are keywords, and importance given by total scores. On the left side of the title item, an envelope mark indicates whether it has been read or not read. The unread article summaries 94-1 and 94-2 are displayed in different colors according to the importance.

  The related keyword list 90 indicates the number of articles requiring attention, and at the same time, by opening the folder, related keywords based on the contents of the keyword database 86 generated by the related keyword classification unit 84 of FIG. 19 are displayed. ing.

  In this example, “○ murder case”, “○ Taro case” and “A child case” are displayed as related keywords, and a number indicating the number of newly arrived articles is displayed next to each item with []. As a result, the user can easily grasp how many articles requiring attention are present. In this figure, in the related keyword list 90, “Taro” is displayed in the new arrival display 95, and an article summary list 92 for two newly arrived articles related to “Taro” is displayed.

  In the browser 96, for example, the content of the article summary 94-1 having the highest importance is displayed by default. The contents of the browser 96 can be opened by clicking on the envelope mark of the title to access the corresponding URL and display the corresponding document.

  In this example, since the article 94-1 includes a related keyword of “A child case” in addition to “Taro △△”, when this article is read, “A” in the related keyword list 90 is displayed. The number of newly arrived articles in “Child Incident” also decreases by one. In this case, even if “A child case” is selected in the related keyword list 90, the same article 94-1 is displayed in the article summary list 92.

  FIG. 21 is a flowchart showing the latest reputation information notification process for a slander article and a human rights violation article according to the embodiment of FIG. In FIG. 21, the presence or absence of a keyword input by the user is checked in step S1, and when the user inputs a malicious reputation word to be checked, a so-called cautionary word, the keyword list 240 is used as a keyword for each user in step S22. Register with.

  FIG. 22 shows the contents of the keyword list 240 of FIG. The keyword list 240 includes an ID and a keyword requiring attention. As the caution keywords, caution reputation words that are malicious reputation words such as “criminals” and “leave” are registered.

  Referring to FIG. 21 again, when the keyword registration is completed, the process proceeds to step S3. When it is determined that the polling unit 28 has reached a certain polling cycle, the process proceeds to step S4 and the latest information search unit 26 is used. Then, a document on the Internet 12 is searched for a cautionary reputable word that is a registered keyword.

  Subsequently, reputation information extraction processing is executed by the reputation information extraction unit 30 in step S5. The details of the reputation information extraction process in step S5 are basically the same as those in the first embodiment shown in the flowchart of FIG. 7, but in the reputation pair extraction process shown in the flowchart of FIG. Since the reputation word is registered as a keyword in advance, the process of extracting the reputation word from the search document is not necessary, and it is different in that the object only needs to be extracted.

  FIG. 23 is a flowchart showing details of the reputation pair extraction process performed in the reputation information extraction process of step S5 of FIG. 21, and corresponds to the reputation pair extraction process of FIG. 8 in the embodiment of FIG.

  In the reputation pair extraction process of FIG. 23, after the morphological analysis of the search document is performed in step S1, the object is extracted from the specific expression and the noun phrase in step S2. Subsequently, in step S3, since the reputable cautionary word is registered as a keyword in advance when checking the slander document or human rights violation document, it is not necessary to extract the reputable word from the search document. Reputation pairs are extracted by combining the extracted target objects with cautionary reputable words registered as keywords.

  Subsequently, in step S4, the reputation pair syntax score C4 is determined from the feature amount in the reputation pair search document, for example, the distance between the two, and then extracted in the search document reputation pair database 34 and the search document reputation word database 36 in step S5. Store information.

  FIG. 24 shows the contents of the search document reputation pair database 34 shown in FIG. 19, which is composed of ID, URL, reputation pair, positive / negative, and syntax score. Of these, as for the reputation pair, for example, for ID = 001, “Taro, criminal” is stored, and the reputable cautionary word registered as a keyword is “criminal”, and the object extracted from the search document Is “Taro”.

  FIG. 25 is an explanatory diagram showing the contents of the search document reputation word database 36 of FIG. 19, and is composed of ID, URL, reputation word, word score, positive / negative, and appearance frequency. Here, the reputation word “criminal” is a reputation word registered in the keyword list 240 as shown in FIG. 22, and the word score and positive / negative are values obtained from the reputation word dictionary 32 of FIG.

  Referring to FIG. 21 again, if the reputation information extraction process is completed in step S5, the process proceeds to step S6, and the related keyword classification process is executed. The related keyword classification process is realized as a function of the related keyword classification unit 84 of FIG. 19, and extracts co-occurrence words related to the cautionary reputed words registered as keywords from the retrieved documents and registers them in the co-occurrence word database 86. After that, the co-occurrence words registered in the co-occurrence word database 86 by the reputation information display unit 40 are displayed as related keywords, for example, as shown in the related keyword list 90 on the user screen 88 of FIG.

  FIG. 26 is an explanatory diagram showing the contents of the co-occurrence word database 86 of FIG. The co-occurrence word database 86 includes an ID, a URL, a word, and an appearance frequency. In the co-occurrence word database 86 of FIG. 26, “○ Taro”, “A child case”, “○ × murder” are co-occurrence words related to “criminals”, which are reputable caution words registered as keywords. "Incident" is registered, and the appearance frequency of each search document is registered.

  The extraction of the co-occurrence words related to the attention-reputed word is obtained by obtaining a syntax score C1 based on the feature amount for the attention-reputed word “criminal” in the search document and the word obtained by the syntax analysis, A word having the following syntax score C1 may be registered in the co-occurrence word database 86 as a co-occurrence word. In addition to the determination based on the syntax score, a word that becomes a co-occurrence word may be determined by matching using a co-occurrence word dictionary prepared in advance.

  Referring again to FIG. 21, if the co-occurrence word database 86 has been registered by the related keyword classification process in step S6, the process proceeds to step S7, and a reputation notification database 420 is created.

  FIG. 27 is an explanatory diagram showing the contents of the reputation notification database 420 of FIG. In FIG. 27, the reputation notification database 420 includes ID, URL, positive / negative, title, news source, update date, reputation word, and related keyword category.

  Here, URL, positive / negative (always negative in the present embodiment) and reputation word are information obtained by the processing of the reputation information extraction unit 30, and the title, news source, and update date / time are shown in FIG. Similar to the form, the document is extracted from the RDF of the target document. If there is no RDF, the title and update date of the article are extracted and stored using the HTML title tag and the file update date.

  In the reputation notification database 420 of this embodiment, a related keyword category is newly added, and here, words that are co-occurrence words registered in the co-occurrence word database 86 shown in FIG. 26 are stored. Yes.

  Referring to FIG. 21 again, when the reputation notification database creation processing is completed in step S7, the process proceeds to step S8. When the user's browsing request is determined, the reputation information is read from the reputation notification database 420 in step S9, for example, Displayed as shown on the user screen 88 in FIG.

  Further, in step S10, for example, it is checked whether or not it is a batch processing cycle performed once a day at midnight. When the batch processing cycle is determined, a browsing history reflection process is executed in step S11. This browsing history reflection process is the same as that shown in the flowchart of FIG. 16 for the embodiment of FIG. By this browsing history reflection processing, articles with high importance that are not being browsed by the user are displayed at a lower rank after the importance is lowered after updating.

(Addition of flame score calculation above)
Here, as the score assignment process performed by the reputation information extraction processing unit 30 in the embodiment of FIG. 19, in addition to the same processing as the flowchart of FIG. Inappropriate expression in the blog article becomes a problem, criticism comments are concentrated, and there are cases where the above-mentioned state of fire is in the state. In this case, the score for the above-mentioned flame shown in FIG. Appropriate processing can be performed by applying the grant process.

  Further, the present invention provides a latest reputation information notification program executed on a computer, and the programs for realizing the functions shown in the latest reputation information notification device 10 of FIG. 19 are shown in FIGS. 11, FIG. 12 and FIG. 16, and the contents shown in the flowchart of FIG. Further, the program for realizing the function of the latest reputation information notification device 100 in the embodiment of FIG. 19 has the contents shown in the flowcharts of FIGS.

  Furthermore, the present invention provides a recording medium storing a program for realizing the functions of the latest reputation information notification device 10 of FIG. 1 and the latest reputation information notification device 100 of FIG. This recording medium is a portable storage medium such as a CD-ROM, floppy disk (R), DVD disk, magneto-optical disk, IC card, etc., a storage device such as a hard disk drive provided inside or outside the computer system, and a line. Including a database or other computer system that holds the program via the network, the database, and a transmission medium on the line.

  In the above embodiment, the case where a document existing on the Internet is searched and the latest information is notified is taken as an example. However, the present embodiment is not limited to a document on the Internet, and is used in a company or on an appropriate network. It can be similarly applied to documents existing in.

  Further, the present invention is not limited to a network, and can be applied to notification of the latest information in which an object or a reputation word is designated by a keyword for a document stored in a database or the like.

  Further, the present invention includes appropriate modifications that do not impair the object and advantages thereof, and is not limited by the numerical values shown in the above embodiments.

Here, the features of the present invention are enumerated as follows.
(Appendix)

(Appendix 1) (Program)
Computer
A keyword registration section that registers objects that you want to know about as a keyword,
A search unit that periodically searches for documents existing on the Internet for the keywords,
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. And a reputation information extraction unit for registering document summary information together with the score,
Reputation information display unit for displaying summary information of documents in the database ranked by the score,
The latest reputation information notification program characterized by functioning as (1)

(Appendix 2) (Details of search part)
In the latest reputation information notification program according to attachment 1, the search unit includes:
The latest information search part that searches the latest information by visiting the information sources on the Internet,
A polling unit that periodically operates the latest information search unit to search for a document including the keyword;
The latest reputation information notification program characterized by comprising.

(Appendix 3) (Details of reputation information extraction)
In the latest reputation information notification program according to attachment 1, the reputation information extraction unit includes:
A reputation pair generation unit that extracts one or more reputation words from the retrieved documents and generates a reputation pair that is combined with each of the objects registered as the keywords;
A score calculator for calculating the score of one or more reputation pairs included in the document;
A database registration unit for registering summary information including an information source address (URL) of a document, an identifier indicating a positive reputation or a negative reputation, a title, and an information source name together with the reputation word and score;
The latest reputation information notification program characterized by comprising. (2)

(Appendix 4) (Score calculation details)
In the latest reputation information notification program according to attachment 1, the score calculation unit obtains, for each document, a syntax score, a word score, an appearance frequency score, and an information source score for each reputation pair in the document, and based on this The latest reputation information notification program characterized by calculating the total score used for ranking. (3)

(Appendix 5) (Score details)
In the latest reputation information notification program described in Appendix 4,
The syntax score is a value indicating the probability of the reputation pair calculated based on the feature amount in the document of the object and reputation word constituting the reputation pair,
The word score is a value indicating the strength of expression of a reputation word set in advance,
The appearance frequency score is a value indicating the number of appearances of reputation words in the document,
The latest reputation information notification program, wherein the information source score is a value representing the reliability of the document as an information source.

(Appendix 6) (Browsing history reflected)
In the latest reputation information notification program according to appendix 4, the computer is further caused to function as a browsing history reflecting unit that updates the information source score using a browsing history of a document on a network within a certain period. The latest reputation information notification program.

(Appendix 7) (Reflecting browsing history)
In the latest reputation information notification program according to attachment 6, the browsing history reflection unit extracts, from the browsing history, information source document addresses that have not been registered in the search destination until then for a certain period of time. The latest reputation information notification program characterized by being added to a new patrol destination.

(Appendix 8) (Details of reputation display)
In the latest reputation information notification program according to appendix 1, the reputation display unit ranks the summary information of the documents in the database according to the score, displays the new arrival, and manages unread and read. Reputation information notification program.

(Appendix 9) (Recording medium)
Computer
A keyword registration section that registers objects that you want to know about as a keyword,
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. And a reputation information extraction unit for registering document summary information together with the score,
Reputation information display unit for displaying summary information of documents in the database ranked by the score,
A computer-readable recording medium storing a latest reputation information notification program that functions as a storage medium.

(Appendix 10) (Device)
A keyword registration unit that registers the target object of the reputation as a keyword,
A search unit that periodically searches for documents existing on the Internet for the keyword;
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. A reputation information extraction unit for registering document summary information together with a score;
Reputation information display unit that displays summary information of documents in the database ranked by the score;
The latest reputation information notification device characterized by comprising: (4)

(Appendix 11) (Method)
A keyword registration step for registering an object whose reputation is desired as a keyword,
A search step for periodically searching for documents existing on the Internet for the keyword;
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. And a reputation information extraction step for registering document summary information together with the score;
Reputation information display step for displaying summary information of documents in the database ranked according to the score;
The latest reputation information notification method characterized by comprising: (5)

(Supplementary note 12) (Notice expression to be noted: Scene B: Latest reputation information notification program)
Computer
A keyword registration section that registers reputable words as keywords,
A search unit for searching for documents existing on the Internet for the keyword;
Analyzing the document searched by the search unit to extract a reputation pair that combines a reputable word of interest and an object registered as the keyword and calculating a score indicating the importance of the reputation pair, Reputation information extractor for registering summary information of documents together with reputable vocabulary and scores,
Reputation information display unit for displaying summary information of documents in the database ranked by the score,
The latest reputation information notification program characterized by functioning as (6)

(Supplementary note 13) (Related keyword classification)
In the latest reputation information notification program according to appendix 12, the computer further extracts a co-occurrence word related to the reputable caution word registered as the keyword from the retrieved document, and registers it in the database. The latest reputation information notification program, which causes the reputation information display unit to function as a related keyword classification unit that displays the co-occurrence words as related keywords together with the summary information. (7)

(Appendix 14) (Details of the search part)
In the latest reputation information notification program according to attachment 12, the search unit includes:
The latest information search part that searches the latest information by visiting the information sources on the Internet,
A polling unit that periodically operates the latest information search unit to search for a document including the keyword;
The latest reputation information notification program characterized by comprising.

(Appendix 15) (Details of reputation information extraction)
In the latest reputation information notification program according to attachment 12, the reputation information extraction unit includes:
A reputation pair generation unit that generates a reputation pair by extracting and combining objects corresponding to the reputable word of interest registered as the keyword from the retrieved document;
A score calculator for calculating a score based on a reputation pair of the document;
A database registration unit for registering summary information including the score, an information source address of the document, an identifier indicating whether the reputation is positive or negative, a title, and an information source name in the database;
The latest reputation information notification program characterized by comprising. (8)

(Appendix 16) (Score calculation details)
In the latest reputation information notification program according to attachment 12, the score calculation unit obtains a syntax score, a word score, an appearance frequency score, and an information source score for each reputation pair, and uses them for the ranking based on the score. The latest reputation information notification program characterized by calculating the total score.

(Supplementary Note 17) (Score Details)
In the latest reputation information notification program described in Appendix 12,
The syntax score is a value indicating the certainty of the reputation pair calculated based on the feature amount in the document of the reputable attention word constituting the reputation pair,
The word score is a value indicating the strength of a precautionary reputation word set in advance,
The appearance frequency score is a value indicating the number of occurrences of cautionary reputation words in the document,
The latest reputation information notification program, wherein the information source score is a value representing the reliability of the document as an information source.

(Appendix 18) (Reflecting browsing history)
The latest reputation information notifying program according to appendix 17, further comprising a browsing history reflecting unit for updating the information source score using a browsing history of a document on a network at regular intervals. Reputation information notification program.

(Appendix 19) (Reflecting browsing history)
In the latest reputation information notification program according to attachment 16, the browsing history reflection unit extracts, from the browsing history, information source document addresses that have not been registered in the search destination until then for a certain period of time. The latest reputation information notification program characterized by being added to a new patrol destination.

(Supplementary note 20) (Notice of caution expression: Scene B: recording medium)
Computer
A keyword registration section that registers reputable words as keywords,
A search unit for searching for documents existing on the Internet for the keyword;
Analyzing the document searched by the search unit to extract a reputation pair that combines a reputable word of interest and an object registered as the keyword and calculating a score indicating the importance of the reputation pair, Reputation information display section for registering summary information of documents together with reputable vocabulary and scores,
A computer-readable recording medium storing a latest reputation information notification program that functions as a storage medium.

(Supplementary Note 21) (Notice of Caution Expression: Scene B: Device)
A keyword registration section that registers reputable words as keywords,
A search unit for searching a document existing on the Internet for the keyword;
Analyzing the document searched by the search unit to extract a reputation pair that combines a reputable word of interest and an object registered as the keyword and calculating a score indicating the importance of the reputation pair, A reputation information extraction unit for registering summary information of a document together with a reputable word of interest and score;
Reputation information display unit that displays summary information of documents in the database ranked by the score;
The latest reputation information notification device characterized by comprising:

(Supplementary note 22) (Notice expression requiring attention: Scene B: Method)
A keyword registration step for registering reputable words as keywords,
A search step of searching for documents existing on the Internet for the keyword;
Analyzing the document searched in the search step to extract a reputation pair that combines a reputable word of interest and an object registered as the keyword, calculate a score indicating the importance of the reputation pair, and store the score in the database. Reputation information extraction step for registering summary information of a document together with reputable vocabulary words and scores;
Reputation information display step for displaying summary information of documents in the database ranked according to the score;
The latest reputation information notification method characterized by comprising:

The block diagram of the functional structure which showed embodiment of the latest reputation information notification apparatus by this invention A block diagram showing a hardware environment of a computer on which the latest information notification program according to the present embodiment is executed Explanatory drawing showing the user screen displaying the latest reputation information notification Explanatory drawing showing another example of the user screen on which the latest reputation information notification is displayed The flowchart which showed the latest reputation information notification process by embodiment of FIG. Explanatory drawing showing the contents of the keyword list of FIG. The flowchart which showed the detail of the reputation information extraction process in step S5 of FIG. The flowchart which showed the detail of the reputation pair extraction process in step S2 of FIG. Explanatory drawing showing the contents of the search document reputation database of FIG. Explanatory diagram showing the contents of the search document reputation word database of FIG. The flowchart which showed the detail of the score provision process in FIG.7 S4 The flowchart which showed the detail of the information source score calculation process in step S4 of FIG. Explanatory drawing which showed the information source score counter table stored in the information source score counter of FIG. Explanatory drawing which showed the information source score counter table stored in the information source score counter of FIG. Explanatory diagram showing the contents of the reputation notification database of FIG. The flowchart which showed the detail of the browsing history reflection process in step S10 of FIG. Explanatory diagram showing the contents of the browsing history database of FIG. The flowchart which showed other embodiment of the score provision process in step S4 of FIG. Functional configuration block diagram showing another embodiment of the latest reputation information notifying device according to the present invention for retrieving the latest article of attention-reputed word Explanatory drawing which showed the user screen of the latest reputation information notification displayed by embodiment of FIG. The flowchart which showed the latest reputation information notification process by embodiment of FIG. Explanatory drawing showing the contents of the keyword list of FIG. The flowchart which showed the detail of the reputation pair extraction process performed in the reputation information extraction process of step S5 of FIG. Explanatory diagram showing the contents of the search document reputation database of FIG. Explanatory drawing showing the contents of the search document reputation word database of FIG. Explanatory diagram showing the contents of the co-occurrence word database of FIG. Explanatory drawing showing the contents of the reputation notification database of FIG.

Explanation of symbols

10, 100: Latest reputation information notification device 12: Internet 14: Blog site 16: News site 18: Review site 20: Bulletin board site 22: Keyword registration unit 24, 240: Keyword list 26: Latest information search unit 28: Polling unit 30 : Reputation information extraction unit 32: Reputation word dictionary 34: Retrieval document reputation vs database 36: Retrieval document reputation word database 38: Information source score counter 40: Reputation information display units 42 and 420: Reputation notification database 44: Browsing history database 46 : Browsing history reflection unit 48: CPU
50: Bus 52: RAM
54: ROM
56: hard disk drive 58: device interface 60: keyboard 62: mouse 64: display 66: network adapter 70, 82, 88: user screen 72: keyword list 74, 95: new arrival display 75, 92: article summary list 76-1, 76-2, 94-1 and 94-2: article summary 78, 96: browser 80: marking 84: related keyword classification unit 86: co-occurrence word database 90: related keyword list

Claims (9)

Computer
A keyword registration section that registers objects that you want to know about as a keyword,
A search unit that periodically searches for documents existing on the Internet for the keywords,
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. And a reputation information extraction unit for registering document summary information together with the score,
Reputation information display unit for displaying summary information of documents in the database ranked by the score,
The latest reputation information notification program characterized by functioning as
The latest reputation information notification program according to claim 1, wherein the reputation information extraction unit includes:
A reputation pair generation unit that extracts one or more reputation words from the retrieved documents and generates a reputation pair that is combined with each of the objects registered as the keywords;
A score calculator for calculating the score of one or more reputation pairs included in the document;
A database registration unit for registering summary information including an information source address (URL) of a document, an identifier indicating a positive reputation or a negative reputation, a title, and an information source name together with the reputation word and score;
The latest reputation information notification program characterized by comprising.
2. The latest reputation information notification program according to claim 1, wherein the score calculation unit obtains, for each document, a syntax score, a word score, an appearance frequency score, and an information source score for each reputation pair in the document, based on the score. The latest reputation information notification program characterized by calculating the total score used for ranking.
In the latest reputation information notification program according to claim 3,
The syntax score is a value indicating the probability of the reputation pair calculated based on the feature amount in the document of the object and reputation word constituting the reputation pair,
The word score is a value indicating the strength of expression of a reputation word set in advance,
The appearance frequency score is a value indicating the number of appearances of reputation words in the document,
The latest reputation information notification program, wherein the information source score is a value representing the reliability of the document as an information source.
A keyword registration unit that registers the target object of the reputation as a keyword,
A search unit that periodically searches for documents existing on the Internet for the keyword;
The document searched by the search unit is analyzed to extract a reputation pair that combines the object registered as the keyword and a reputation word, and a score indicating the importance of the reputation pair is calculated, and the reputation word is stored in a database. A reputation information extraction unit for registering document summary information together with a score;
Reputation information display unit that displays summary information of documents in the database ranked by the score;
The latest reputation information notification device characterized by comprising:
A method executed by a computer including a keyword registration unit, a search unit, a reputation information extraction unit, and a reputation information display unit,
A keyword registration step in which the keyword registration unit registers an object whose reputation is to be known as a keyword;
A search step in which the search unit periodically searches for documents existing on the Internet for the keyword;
The reputation information extraction unit analyzes the document searched by the search unit, extracts a reputation pair that combines the object registered as the keyword and a reputation word, and obtains a score indicating the importance of the reputation pair A reputation information extraction step of calculating and registering document summary information together with the reputation word and score in a database;
The reputation information display unit displays summary information of documents in the database ranked by the score and displayed, and
The latest reputation information notification method characterized by comprising: Computer
A keyword registration section that registers reputable words as keywords,
A search unit for searching for documents existing on the Internet for the keyword;
Analyzing the document searched by the search unit to extract a reputation pair that combines a reputable word of interest and an object registered as the keyword and calculating a score indicating the importance of the reputation pair, Reputation information extractor for registering summary information of documents together with reputable vocabulary and scores,
Reputation information display unit for displaying summary information of documents in the database ranked by the score,
The latest reputation information notification program characterized by functioning as
8. The latest reputation information notification program according to claim 7 , wherein the computer further extracts a co-occurrence word related to a reputable caution word registered as the keyword from a retrieved document and registers the extracted word in the database. The latest reputation information notification program causing the reputation information display unit to function as a related keyword classification unit that displays the co-occurrence word as a related keyword together with the summary information.
The latest reputation information notification program according to claim 7 , wherein the reputation information extraction unit includes:
A reputation pair generation unit that generates a reputation pair by extracting and combining objects corresponding to the reputable word of interest registered as the keyword from the retrieved document;
A score calculator for calculating a score based on a reputation pair of the document;
A database registration unit for registering summary information including the score, an information source address of the document, an identifier indicating whether the reputation is positive or negative, a title, and an information source name in the database;
The latest reputation information notification program characterized by comprising.

Images