JP2009116457A - Method and device for analyzing internet site information - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an analysis device and an analysis method for effectively and accurately recognizing meanings, backgrounds and trends included in a large number of items of site information disclosed on the Internet on the basis of the site information. <P>SOLUTION: The device includes: a crawler 14 that is an information collection means for collecting text information of Web sites 12; a means for dividing words through analyzers 18 dividing the text information into words; an evaluation information extraction means for extracting evaluation information composed of evaluation objects, evaluation axes showing characteristics of the evaluation objects and evaluation expression from groups of the words through the analyzers 18; an evaluation classification means by evaluation axes through the analyzers 18 classifying the evaluation information for each evaluation axis; an evaluation expression dictionary group database 20 with a plurality of evaluation expression dictionaries composed of evaluation expression and scores showing its extent; an evaluation score calculation means that are the analyzers 18 calculating the score for each item of the evaluation information; and an evaluation analysis output means composed of a portal server 24 relatively comparing to output the scores of any different evaluation axes. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to an Internet site information analysis method and apparatus for analyzing information published on an Internet website and acquiring / providing trend information and the like.

On the Internet, an enormous amount of information released by various people is stored in a Web site and is still increasing. Here, the website is a bulletin board (BBS: Bulletin Board).
System), homepage, blog (Web) called blog
Log).

  In recent years, efforts to obtain new trend information by analyzing articles accumulated on Web sites have been actively conducted. For example, a reputation analysis technique that analyzes how positive or negative the content of an opinion is, a technique that evaluates trends in the frequency of appearance and the degree of attention (Burst degree) of a given keyword in time series, Services using various technologies are actually provided, such as a search engine optimization technology that allows a specific Web site to be displayed at the top of search results when a search engine is used.

  As a specific example, as disclosed in Patent Document 1, a plurality of pieces of text information are acquired from a website, and from each piece of text information, an evaluation target such as an evaluation axis indicating the nature of the evaluation target, , Extract reputation information consisting of a set of evaluation expression representing evaluation for the evaluation axis and an evaluation score set in advance in the evaluation expression, and aggregate all evaluation scores for each reputation information by a predetermined method (summation) By doing so, there is a reputation information processing apparatus that obtains knowledge about the subject matter to be analyzed. In this embodiment of the reputation information processing apparatus, specifically, the evaluation score for the evaluation expression is 0.2 for “cool”, 0.3 for “light”, and 0.1 for “good”. And an analysis method for quantifying the degree of negativity / affirmation of reputation information.

  Moreover, as disclosed in Patent Document 2, a plurality of pieces of text information are acquired from a website, reputation information with an evaluation score for each reputation is acquired, and an evaluation score is assigned to each reputation information. There is a reputation information processing apparatus that obtains knowledge about a subject matter that is an analysis target by tabulating with a calculation method. In particular, if you set a predetermined weighting value for each evaluation axis such as “price” and “design”, for example, when the evaluation axis “design” is attracting attention in the world, reputation information about “design” Even if the number is small, it is devised so that the evaluation score related to “design” is largely reflected in the total result.

Furthermore, as disclosed in Patent Document 3, the website information is accessed, text information (reputation information) is collected every predetermined period, the usage status of the collected keywords is quantified, and the quantified keyword By monitoring usage, select keywords that will become trend keywords from the extracted keywords, predict trend keywords that are likely to be used in search engines in the near future, and relate to those trend keywords There is a trend prediction device that provides information to be used. This is characterized in that it has better real-time performance than the trend keyword is obtained by analyzing the actual use of the search keyword input to the search engine.
JP 2007-172051 A JP 2006-53685 A JP 2006-227965 A

  However, in the reputation information processing apparatus of Patent Document 1, the evaluation expressions such as “cool”, “light”, and “good” and the evaluation scores set for each of the evaluation expressions are, for example, “camera” and “battery possession”. It is also used in common when evaluating different evaluation axes. Therefore, it was impossible to perform appropriate evaluation according to each evaluation axis. In addition, since the evaluation scores of a lot of reputation information are finally handled as one evaluation expression and a comprehensive evaluation value that is aggregated for each evaluation axis, it can be used for different evaluation axes that one contributor has. The correlation of opinions is buried in the comprehensive evaluation value. Therefore, for example, even if there are articles from two contributors, such as “I feel good because the design is good” and “I feel that the design is good, but the price is high”, what kind of priority is given to each contributor? I couldn't go into the analysis of how to measure the degree or whether there is a tendency to balance.

  In the reputation information processing apparatus of Patent Document 2, the device has been devised so that the value of each evaluation information is not buried in the comprehensive evaluation value obtained by collecting the evaluation scores. As in the case of No. 1, it was impossible to analyze a plurality of requests possessed by each contributor, including how to assign priorities and a tendency toward balance.

  In the trend prediction device of Patent Document 3, the affirmative / negative of text information including the keyword and the frequency of occurrence of the keyword are analyzed for the keyword to be analyzed, and the keyword itself is used as the trend keyword. It was a decision whether or not to select, and it was not possible to substantially grasp the change in the interest of the contributor spreading around the keyword.

  Further, a correlation analysis method has been proposed as a method of collecting peripheral information of a specific keyword, and a service such as an associative search for acquiring another keyword associated with the keyword has been performed. Similarly, changes in the interests of contributors spreading around the keyword could not be grasped in practice.

  In addition, there is a request to know in real time an active website where useful information can be obtained, but no method or service that can meet such a request has been proposed and has not yet been put into practical use. .

  The present invention was made in view of the above background art, and based on a lot of site information published on the Internet, an analysis that can effectively and accurately know the meaning, background, and tendency of the information. Providing a device and an analysis method, in particular, a reputation analysis that evaluates a plurality of requests of one article in comparison, a website activity analysis that extracts an active website that can provide useful information in real time, Internet site information analysis method that can analyze the co-occurrence information that actually grasps the change of the interest of the poster that spreads around a certain keyword, and can know the new trend accurately from the articles accumulated on the website And to provide a device.

The present invention relates to an Internet site information analysis method for accessing and analyzing the website information existing on the Internet, collecting the sentence information, an information collecting step for collecting the sentence information of the website, and the sentence information. A word dividing step for dividing a word into words, and a reputation that extracts from the word group a reputation information consisting of a set of an evaluation object, an evaluation axis representing the nature of the evaluation object, and an evaluation expression representing an evaluation with respect to the evaluation axis An information extraction step;
A plurality of evaluation expression dictionaries are provided in advance for each evaluation axis, and a plurality of evaluation expression dictionaries are provided for each evaluation axis. An internet site comprising: a score calculation step for calculating a score based on an evaluation expression dictionary corresponding to the evaluation axis; and a reputation analysis output step for comparing and outputting scores of arbitrary different evaluation axes for each sentence information Information analysis method.

  Further, the present invention provides an information collecting step for collecting the sentence information and update date information of each website in an internet site information analyzing method for collecting and analyzing sentence information by accessing a website existing on the Internet. A word dividing step for dividing the sentence information into words, and extracting a word that is the same as or similar to a predetermined keyword from the word group and calculating the number of pieces of related information that includes the word A related information posting rate calculating step for calculating the ratio of the number of related information posts to the number of sentence information collected from each website, for each website, a reference date for performing analysis, and the update date information In addition, an update frequency calculation step for calculating the update frequency of each website, the related information posting rate for each website, and the update An Internet site information analyzing method and a site activity analysis output step of outputting the frequency relative comparison.

  Further, according to the present invention, in the Internet site information analysis method for accessing and analyzing the website information on the Internet, collecting the sentence information, the collection of the sentence information of the website is performed every time a predetermined period elapses. An information collecting step for performing, a word dividing step for dividing the sentence information into words, an investigation target information extracting step for extracting sentence information including words that are the same as or similar to the keyword to be investigated from the word group, Based on the co-occurrence keyword extraction step for extracting a co-occurrence keyword that is included in the words constituting the investigation target information and is a keyword different from the keyword, and the frequency of occurrence of the co-occurrence keyword in the investigation target information, A co-occurrence keyword score calculating step for calculating a score for each co-occurrence keyword; A sorting step for rearranging the keyword occurrence scores to create a co-occurrence information list and a co-occurrence information analysis output step for outputting the co-occurrence information list obtained for each predetermined period in time series are provided. Internet site information analysis method.

  Further, the present invention is an information that is constituted by a computer system, collects text information by accessing a web site existing on the Internet, collects text information, and analyzes the text information of the web site. From a set of a collecting means, a word dividing means for dividing the sentence information into words, an evaluation object, an evaluation axis representing the property of the evaluation object, and an evaluation expression representing an evaluation with respect to the evaluation axis, from the word group A reputation information extracting means for extracting reputation information, an evaluation axis-specific classification means for classifying the reputation information for each evaluation axis, and an evaluation expression dictionary comprising a set of the evaluation expression and a score indicating the degree, A plurality of evaluation expression dictionary group databases provided for each evaluation axis and an evaluation expression dictionary group data corresponding to the evaluation axis to be analyzed A score calculation means for calculating a score for each reputation information based on the evaluation expression dictionary, and a reputation analysis output means for comparing and outputting scores of arbitrary different evaluation axes for each sentence information Is an Internet site information analyzing apparatus.

  The reputation analysis output means outputs a two-dimensional graph based on the scores of two different evaluation axes or a three-dimensional graph based on the scores of three different evaluation axes.

  Further, the present invention is an information that is constituted by a computer system, collects text information by accessing a web site existing on the Internet, collects text information, and analyzes the text information of the web site. A collecting means; a word dividing means for dividing the sentence information into words; and a related information post for extracting a word that is the same as or similar to a predetermined keyword from the word group and calculating the number of sentence information including the word A number calculating means, a related information posting rate calculating means for calculating, for each website, a ratio of the number of related information posts to the number of text information collected from each website, and an update for collecting an update date of each website. Update that calculates the update frequency of each website based on the date information collection means, the reference date for analysis, and the update date information A degree calculating unit, an Internet site information analyzer and a site activity analysis output means for outputting said update frequency and the related information posts index for each Web site by relative comparison.

  The activity analysis output means outputs the related information posting rate and the update frequency in a two-dimensional graph.

  Further, according to the present invention, in an Internet site information analyzing apparatus that accesses and analyzes a website on the Internet, collects the sentence information, and collects the sentence information of the website every time a predetermined period elapses. Information collecting means for performing, word dividing means for dividing the sentence information into words, investigation target information extracting means for extracting sentence information including words that are the same as or similar to the keyword to be investigated from the word group, and A co-occurrence keyword extracting means for extracting a co-occurrence keyword that is included in words constituting the investigation target information and is other words excluding the same or similar word as the investigation target keyword; and the co-occurrence keyword in the investigation target information A co-occurrence keyword score calculating means for calculating a score for each co-occurrence keyword based on the frequency of occurrence of the co-occurrence keyword, and the co-occurrence keyword Sorting means for rearranging the keywords in the order of the co-occurrence keyword scores to create a co-occurrence information list, and co-occurrence information analysis output for outputting the co-occurrence information list obtained for each predetermined period in time series And an Internet site information analyzing apparatus.

  According to the present invention, it is possible to analyze an enormous amount of information released and accumulated on a website and easily obtain accurate trend information.

  In particular, according to the inventions of claims 1, 4 and 5, it is possible to accurately calculate the score of the contributor's opinion by providing an evaluation expression dictionary and an evaluation score for each evaluation axis. It is possible to perform reputation analysis that goes into the way of assigning priorities to the evaluation axes (requests) of each person and the tendency of balance.

  Further, according to the inventions of claims 2, 6 and 7, the website actively transmitting information by the site activity analysis for calculating the posting rate and the update frequency of the information related to the predetermined keyword for each website. If information is collected only on Web sites that should be noted as information sources, useful trend information can be obtained efficiently.

  Furthermore, according to the third and eighth aspects of the present invention, the contributor's information spreading around the predetermined keyword is analyzed by co-occurrence information analysis that analyzes the change of another keyword co-occurring with the predetermined keyword in time series. The change in interest can be grasped in practice.

  Hereinafter, an embodiment of a network system in which the Internet site information analysis apparatus 10 of the present invention is arranged will be described with reference to FIG. In this network system, a crawler 14 that regularly collects text information in RSS (Rich Site Summary) format from a designated website and a website 12 on which many people publish information such as comments and opinions. An article database 16 in which information collected by the crawler 14 is stored as a database, an analyzer 18 having a program for performing three analyzes described later, an evaluation expression corresponding to the evaluation axis to be analyzed, and its evaluation An evaluation expression dictionary group database 20 in which scores are set, an analysis result database 22 for storing analysis results, and a personal computer owned by a user 26 are connected, and desired analysis results are extracted from the analysis result database 22. The portal server 24 distributed to the user 26 is arranged on the Internet. To have.

  An Internet site information analysis method for performing reputation analysis according to the first embodiment of the present invention will be described with reference to FIGS. First, an outline will be described based on the flow shown in FIG. In step S110, sentence information is collected from each website as a blog. In the Internet site information analysis apparatus 10, the crawler 14, the article database 16, and the reputation analysis analyzer 18a function as information collection means. Next, in step S120, the collected sentence information is broken down into words (parts of speech). In the Internet site information analysis apparatus 10, the reputation analysis analyzer 18a functions as the word dividing means. In step S130, reputation information comprising a set of an evaluation object, an evaluation axis representing the nature of the evaluation object, and an evaluation expression representing an evaluation with respect to the evaluation axis is extracted from the divided word group. In the Internet site information analysis apparatus 10, the evaluation analysis analyzer 18a functions as this reputation information extraction means.

  In step S140, the extracted reputation information is classified for each evaluation axis. In the Internet site information analysis apparatus 10, the evaluation analysis analyzer 18a functions as the evaluation axis-specific classification means. In step S150, the evaluation expression dictionary corresponding to the evaluation axis to be analyzed is extracted from the evaluation expression dictionary group database 20, the score of each piece of reputation information is calculated, and the calculation result is stored. In the Internet site information analysis apparatus 10, the evaluation analysis analyzer 18a and the reputation analysis result database 22a serve as the score calculation means. In step S160, in response to a request from the user 26, each sentence information having a score based on any different evaluation axis is displayed on a graph and output. In the Internet site information analysis apparatus 10, the reputation analysis result database 22a and the reputation analysis display framework 24a provided in the portal server 24 function as this reputation analysis output means.

  Next, each step of the reputation analysis will be described in detail. As shown in FIG. 3, each Web site has a plurality of pieces of text information. In step S110, for example, the sentence information divided for each predetermined sentence unit is collected, such as sentence information a1 “the car of company A is expensive and the handle is difficult to use”. Further, these pieces of text information are decomposed into words (parts of speech) such as nouns, particles, adjectives and verbs in step S120.

  In step S130, the words are related to each other using Japanese morphological analysis technology, dependency analysis technology, and the like, which are known techniques. According to the Japanese morphological analysis technology, by analyzing the context in which nouns and particles appear in a sentence, for example, if the word is included in the sentence information a1, “price” corresponds to “high”. , “Handle” corresponds to “Difficult to use”. In this way, the evaluation information that is a set of the evaluation object, the evaluation axis representing the property of the evaluation object, and the evaluation expression representing the evaluation with respect to the evaluation axis is extracted from the text information. For example, in the case of the text information a1, two pieces of reputation information “car, car price, high” and “car, car operability, difficult to use” are extracted.

  In step S140, the reputation information is classified for each evaluation axis such as “car price” and “car operability”. This is a preparatory step for organizing information in order to efficiently perform the calculation processing in step S150 described later.

  In step S150, a score of each piece of reputation information is calculated using a plurality of evaluation expression dictionaries provided in the evaluation expression dictionary group database 20. Here, an example of the evaluation expression dictionary will be described with reference to FIG. The evaluation expression dictionary is provided for each evaluation axis. For example, the evaluation expression dictionary A corresponds to the evaluation axis of “car price”, and it relates to the price of the car such as “high”, “expensive”, “moderate”. Possible evaluation expressions are listed, and a positive / negative level is defined as an evaluation score for each evaluation expression. On the other hand, the evaluation expression dictionary B corresponds to the evaluation axis of “car operability”, and another evaluation expression suitable for evaluating “car operability” is listed, and for each evaluation expression, Positive / negative levels are defined as evaluation scores.

  A more detailed process of step S150 using the evaluation expression dictionary will be described with reference to FIG. Step S151 is a step of paying attention to one evaluation expression dictionary in the evaluation expression dictionary group database 20 and extracting a reputation information group corresponding to the evaluation expression dictionary. For example, when attention is paid to the evaluation expression dictionary A, reputation information in which the evaluation axis of reputation information corresponds to “car price” is extracted.

  In step S152, reputation information including evaluation expressions listed in the evaluation expression dictionary is extracted in order, and in step S153, a calculation is performed to assign a score of the evaluation expression as a score of the reputation information, and the result is obtained. Store in the reputation analysis result database 22a. For example, a calculation is performed in which an evaluation score of “+1” is assigned to reputation information including an evaluation expression “high”, and an evaluation score of “+4” is assigned to reputation information including an evaluation expression “fair”. Save. In step S154, which is a determination step, it is determined whether or not the score calculation has been performed for all evaluation expression dictionaries. If NO, steps S151 to S153 are repeated for the next evaluation expression dictionary, and when YES is obtained. Step S150 ends. Therefore, for example, in the case of the text information a1, in the case of the text information a1, scores for the two evaluation axes “car price = + 1” and “car operability = + 1” are assigned.

  In step S160, the sentence information to which the score is assigned in step S150 is provided as reputation analysis information represented in a graph based on any different evaluation axis. For example, as shown in FIG. 6, when “car price” and “car operability” are selected as evaluation axes, a graph of a scatter diagram is drawn for each piece of text information. This reputation analysis information is provided in a PULL type (the user 26 extracts information as needed). A three-dimensional graph may be drawn when three evaluation axes are selected.

  According to the Internet site information analysis method of reputation analysis (step S110 to step S160) according to the first embodiment of the present invention described above, an evaluation expression dictionary and an evaluation score are provided for each evaluation axis. It is possible to calculate the score by accurately capturing the opinions. In addition, it does not simply summarize the opinions of many evaluation items, but analyzes the meaning of each piece of text information and displays it in a graph with respect to different evaluation axes, so that many unspecified contributors have it. With regard to a large number of requests, it becomes possible to perform a reputation analysis and a trend analysis in which the overall trend is subdivided into more detailed items.

  Next, an Internet site information analysis method for performing site activity analysis according to the second embodiment of the present invention will be described with reference to FIGS. First, an outline will be described based on the flow shown in FIG. In step S210, text information and update date information are collected from each Web site. At this time, in the Internet site information analysis apparatus 10, the crawler 14, the article database 16, and the site activity analysis analyzer 18b function as the information collection unit and the update date information acquisition unit. Next, in step S220, the collected sentence information is decomposed into words (parts of speech). In the Internet site information analyzing apparatus 10, the site activity analysis analyzer 18b functions as this word dividing means. Then, in step S230, the sentence information including the same or similar to the predetermined keyword in the divided word group, that is, the number of related information posts is calculated. In the Internet site information analysis apparatus 10, the site activity analysis analyzer 18b functions as the related information posting number calculation means. In step S240, the related information posting rate, which is the ratio of the number of related information posts to the number of text information collected from the website, is calculated, and the calculation result is stored. In the Internet site information analysis apparatus 10, the site activity analysis analyzer 18b and the site activity analysis result database 22b function as the related information posting rate calculation means.

  Step S250 calculates the update frequency of each Web site based on the reference date for analysis and update date information, and stores the calculation result. The Internet site information analysis apparatus 10 uses the site activity analysis analyzer 18b. The site activity analysis result database 22b serves as the update frequency calculation means.

  In step S260, in response to a request from the user 26, each Web site to which two calculated values of the related information posting rate and the update frequency are assigned is displayed on a graph and output. In the Internet site information analysis apparatus 10, the site activity analysis result database 22b and the site activity analysis display framework 24b provided in the portal server 24 function as this site analysis output means.

  Next, each step of the site activity analysis will be described in detail. As shown in FIG. 8, each Web site includes a plurality of text information and information on the last update date of each Web site. In step S210, for example, for the Web site 1, the text information a1 and a2 and the update date information “last update date: September 11” are collected. Further, in step S220, the sentence information is decomposed into words (parts of speech) such as nouns, adjectives and verbs.

  Further detailed processing of step S230 will be described with reference to FIG. In step S230, when a predetermined keyword to be investigated is given in step S231, a word group similar to the keyword is extracted using a thesaurus which is a kind of thesaurus. In step S232, sentence information including any word group that is the same as or similar to the keyword, that is, related information is extracted for each Web site. In step S233, the extracted related sentence information is counted and the number is cumulatively calculated. In step S234, which is a determination step, it is determined whether or not calculations have been performed for all websites. If NO, steps S232 to S233 are repeated for the next website. End and go to the next step 240.

  Thus, in step S230, for example, when the keyword “car” is given, common words, abbreviations, formal names, and other words such as “light four”, “hybrid car”, and “automobile” are extracted as similar words in step S231. Therefore, it is possible to analyze the contents of related information and calculate the number of posts without omission.

  In step S240, the posting rate of related information is calculated for each Web site, and the calculation result is stored in the site activity analysis result database 22b. The related information posting rate is calculated by dividing the total number of sentence information collected from each Web site by using the denominator and the number of related information for a predetermined keyword in the denominator. For example, the Web site 1 shown in FIG. 8 has two pieces of sentence information a1 and a2, and the related information of the keyword “car” is one piece of sentence information a1 including the word “hybrid car”. Therefore, the related information posting rate for the keyword “car” on the Web site 1 is calculated as 0.5. In this way, the related information posting rate is calculated for each keyword and for each Web site, and the calculation results are systematically stored in the site activity analysis result database 22b as shown in the list of FIG.

  In step S250, the update frequency of each Web site is calculated, and the calculation result is stored in the site activity analysis result database 22b. In the calculation example of FIG. 8, 1 is added to the difference between the reference date for analysis and the last update date of the Web site, and the reciprocal thereof is defined as the update frequency. According to this definition, in the case of Web site 1, the update frequency is 1.0 because both the reference date for analysis and the last update date are September 11 (the same day). In the case of the Web site 2, if the same calculation is performed, the update frequency is 0.011. That is, the frequently updated Web site 1 has a high update frequency, and the Web site 2 that has been left unupdated for a long period shows a low update frequency.

  In step S260, the relative information posting ratio assigned for the predetermined keyword in step S240 is plotted on the horizontal axis, the update frequency assigned in step 250 is plotted on the vertical axis, and the relative positioning of each website is represented in a graph. The site activity analysis information is provided in PULL type (the user 26 extracts information as needed). For example, as shown in FIG. 11, the Web site located in the upper right of the graph is a group of Web sites that have a lot of information related to “cars” and are frequently updated. If you access a website that actively transmits information, you will find that you can get useful information about "cars". Conversely, the website located at the lower left of the graph is a group of websites with little information on “cars” and a low update frequency. Therefore, a website with low activity such as website 8 is accessed. However, it turns out that useful information about "cars" is unlikely to be obtained.

  According to the Internet site information analysis method of the site activity analysis (steps S210 to S260) according to the second embodiment of the present invention described above, it is possible to easily know the website that is actively transmitting information. If information is collected focusing on those Web sites that should be noted as information sources, useful trend information can be obtained efficiently.

  Next, an outline of the Internet site information analysis method for performing co-occurrence information analysis according to the third embodiment of the present invention will be described based on the flow shown in FIGS. In step S310 shown in the flow of FIG. 12, sentence information is collected each time a predetermined period elapses from each Web site. In the Internet site information analyzing apparatus 10, the crawler 14, the article database 16, and the co-occurrence information analyzing analyzer 18c function as this information collecting means. Next, in step S320, the collected sentence information is broken down into words (parts of speech). In the Internet site information analyzing apparatus 10, the co-occurrence information analyzing analyzer 18c functions as the word dividing means. In step S330, sentence information including the same or similar to the predetermined keyword in the divided word group, that is, survey target information is extracted. At this time, in the Internet site information analysis apparatus 10, the co-occurrence information analysis analyzer 18c functions as this investigation target information extraction means. In step S340, other words excluding the same or similar words as the predetermined keyword, that is, co-occurrence keywords are extracted from the words constituting the extracted survey target information. In the Internet site information analysis apparatus 10, the co-occurrence information analysis analyzer 18c functions as this co-occurrence keyword extracting means.

  In step S350, a score is calculated for each co-occurrence keyword based on the frequency with which the extracted co-occurrence keyword appears in the survey target information. In the Internet site information analysis apparatus 10, the co-occurrence information analysis analyzer 18c functions as this co-occurrence keyword score calculation means. Next, in step S360, a co-occurrence information list in which the co-occurrence keyword groups are rearranged in the order of the co-occurrence keyword scores is created and stored. In the Internet site information analysis apparatus 10, the co-occurrence information analysis analyzer 18c and the co-occurrence information analysis result database 22c serve as the sorting means. In step S370, in response to a request from the user 26, the co-occurrence information list created every time a predetermined period elapses is output in a time series. In the Internet site information analysis apparatus 10, the co-occurrence information analysis result database 22c and the co-occurrence information analysis display framework 24c included in the portal server 24 function as the co-occurrence information analysis output means.

  Next, each step of the co-occurrence information analysis will be described in detail. As shown in FIG. 13, each Web site has a plurality of text information. In step S310, for example, if it is Web site 1, sentence information a1 and a2 are collected. Further, in step S320, the sentence information is decomposed into words (parts of speech) such as nouns, adjectives and verbs.

  In step S330, when a predetermined keyword to be investigated is given, a word group similar to the keyword is extracted using a thesaurus which is a kind of thesaurus, and either one of the same or similar word groups as the keyword is extracted. Text information including, that is, survey target information is extracted. For example, according to the example of FIG. 13, when the keyword “digital camera” is given, common names such as “digital camera”, “digital still camera”, and “digital video camera”, abbreviated names, formal names, and other words are extracted as similar words. Then, text information a1, a2, and b2 including “digital camera” and similar words are extracted as investigation target information. In this way, it is possible to extract the investigation object information that should be originally investigated without omission.

  In step S340, other words excluding the same or similar words as the predetermined keyword, that is, co-occurrence keywords, are extracted from the words constituting the extracted survey target information. For example, in the case of sentence information a1, “Company A”, “Company B”, and “Performance” correspond to the co-occurrence keywords. Here, when co-occurrence keywords such as “(A) Co., Ltd.”, “A Co., Ltd.”, “A” and “Company A” are extracted separately, all are synonymous with “Company A”. Collect them as co-occurrence keywords and go to the next step.

  Regarding step S350, further detailed processing will be described based on the flow of FIG. In step S351, the number of all sentence information collected is counted. For example, in the example of FIG. 13, if information collection is performed only on the Web sites 1 and 2, the total number of text information is counted as 5. In step S352, the number of text information corresponding to the survey target information is counted. For example, in the example of FIG. 13, out of the total number 5 of sentence information, the number of pieces of investigation target information for the keyword “digital camera” is counted as 3. In step S353, the number of sentence information including the same word as the co-occurrence keyword extracted in step 340 is counted for each co-occurrence keyword. For example, in the example of FIG. 13, the number of text information including the co-occurrence keyword “Company A” out of the number 5 of text information is counted as 3. In step S354, the number of investigation target information including the same word as the co-occurrence keyword extracted in step S340 is counted for each co-occurrence keyword. For example, in the example of FIG. 13, the number of pieces of investigation target information including the co-occurrence keyword “Company A” is counted as two out of three pieces of investigation target information for “digital camera”.

  In step S355, the score of the co-occurrence keyword is calculated for each co-occurrence keyword. Here, the score of the co-occurrence keyword is obtained by dividing the sum of the count number of S351 by the count number of Step S354 by the sum of the count number of S352 and the count number of S353. It is preferable to define that converted into a logarithm. For example, in the example of FIG. 13, the score of the co-occurrence keyword “Company A” is 5 as the count number in step S351, 3 as the count number in step S352, 3 as the count number in step 353, and the count number in step S354. Is used to obtain a rating of 0.152. In step S356, which is a determination step, it is determined whether or not all co-occurrence keywords have been calculated. If NO, steps S353 to S355 are repeated for the next co-occurrence keyword. S350 ends.

  Note that step S350 of the present embodiment has a feature that the correlation chain element is also incorporated in the score calculation. Originally, when there is a correlation between the keyword “digital camera” and the co-occurrence keyword “Company A”, a keyword group other than “digital camera” co-occurs corresponding to the co-occurrence keyword “A company” It should be considered that there is a correlation between the two. However, if attention is paid to the chain of correlations, there is a problem that the amount of calculation becomes enormous, and generally, the processing up to that point has not been performed. In this embodiment, by incorporating the count number in step S351 and the count number in step S353 into the calculation formula, not only the strength of the correlation between “digital camera” and “Company A” but also “liquid crystal television” and “ Regarding the strength of the correlation of “Company A”, the difference can be relatively reflected in each score.

  In step S360, a co-occurrence information list in which the co-occurrence keyword groups are rearranged in the order of the co-occurrence keyword scores is created and stored in the co-occurrence information analysis result information database 22c. Further, in step S370, the co-occurrence information list created every time a predetermined period elapses in response to a request from the user 26 is displayed in time series, and the PULL type (the user 26 as necessary) Information is taken out). FIG. 15 shows an example in which the score calculation is performed for all websites including the websites 1 and 2. For example, in the case of the keyword “digital camera”, the co-occurrence keyword “product W” had a low score and was not ranked as of July 2007, but it rose to the second place in August 2007. Therefore, it can be seen that in the field of “digital camera”, “product W” has become the center of the contributor's topic. Further, the co-occurrence keyword “Company B” was ranked second in July 2007, but was ranked down to fifth in August 2007. Therefore, it can be seen that the poster has stopped paying attention to “Company B”.

  According to the Internet site information analysis method of the co-occurrence information analysis (steps S310 to S370) according to the third embodiment of the present invention described above, the change of another keyword that co-occurs with a predetermined keyword is time-sequentially. By analyzing, it is possible to grasp in real terms the change in the interest of the poster spreading around the predetermined keyword.

  In addition, this invention is not limited to the said embodiment, The internet site information analysis apparatus 10 is a working means or step concerning any one or more of reputation analysis, site activity analysis, and co-occurrence information analysis. Any analysis device or analysis method may be used.

  Moreover, the display frame of the analysis result output in steps S160 and S260 may be any graph as long as it can visually recognize the relative relationship between a plurality of characteristic values of each data. It is not limited to images. The graph scale should be displayed in logarithmic form, or a legend should be attached to display the results of multiple analysis.

  Further, the calculation formula of the related posting rate defined in step S240, the calculation formula of the update frequency defined in step S250, and the calculation formula of the co-occurrence keyword score defined in step S350 are the items to be investigated and analyzed for each field As long as it is defined in view of individual circumstances, the calculation formula of the above embodiment is not limited. For example, in the case of the co-occurrence keyword score calculation formula defined in step S350, the content to be analyzed in detail, such as changing the base value of the logarithm or substituting a square value for the number of specific information, is the characteristic value. Another calculation formula may be defined so as to appear prominently.

  In addition, a series of processing operations as in the first to third embodiments are constructed as a program, installed in a server computer used as the Internet site information analysis apparatus 10, and executed by a control means such as a CPU. The program may be distributed via a network. Further, the constructed program is stored in a portable storage medium such as a hard disk device, a flexible disk, or a CD-ROM connected to various computers used as the Internet site information analysis apparatus 10, and is installed and executed on the computer. You may do it.

It is a figure which shows the structure of the whole network system which has arrange | positioned one Embodiment of the internet site information analysis apparatus of this invention. It is a flowchart which concerns on reputation analysis which is 1st embodiment of the internet site information analysis method of this invention. It is a figure which shows the example which extracts reputation information from text information in this embodiment. It is a figure which shows the example of the evaluation expression dictionary used in this embodiment. It is a flowchart which concerns on the step which performs a score calculation in this embodiment. It is a graph which shows the example of the output format of a reputation information analysis result in this embodiment. It is a flowchart which concerns on the site activity analysis which is 2nd embodiment of the internet site information analysis method of this invention. It is a figure which shows the example of text information and update frequency calculation in this embodiment. It is a flowchart which concerns on the step which performs the number calculation of related articles in this embodiment. It is a figure which shows the example of the calculation result list | wrist of a related article contribution rate in this embodiment. It is a graph which shows the example of the output format of a site activity analysis result in this embodiment. It is a flowchart which concerns on co-occurrence information analysis which is 3rd embodiment of the internet site information analysis method of this invention. It is a figure which shows the example which extracts investigation object information from text information in this embodiment. It is a flowchart which concerns on the step which performs the score calculation of each co-occurrence keyword in this embodiment. It is a figure which shows the example of the output format of a co-occurrence information analysis result in this embodiment.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Internet site information analyzer 12 Website 14 Crawler 16 Article database 18 Analyzer 20 Evaluation expression dictionary group database 22 Analysis result database 24 Portal server 26 User

Claims (8)

In the Internet site information analysis method for accessing and analyzing the website information on the Internet, collecting the text information,
An information collecting step for collecting text information of the website;
A word dividing step for dividing the sentence information into words;
A reputation information extraction step for extracting reputation information consisting of a set of an evaluation object, an evaluation axis representing the nature of the evaluation object, and an evaluation expression representing an evaluation with respect to the evaluation axis, from the word group;
A classification step for each evaluation axis for classifying the reputation information for each evaluation axis;
A plurality of evaluation expression dictionaries consisting of a set of the evaluation expression and a score indicating its degree are provided in advance for each evaluation axis, and a score calculation for calculating a score based on the evaluation expression dictionary corresponding to the evaluation axis to be analyzed therein Steps,
For each sentence information, a reputation analysis output step for outputting a relative comparison of scores of arbitrary different evaluation axes;
An Internet site information analysis method comprising: In the Internet site information analysis method for accessing and analyzing the website information on the Internet, collecting the text information,
An information collecting step of collecting the sentence information and update date information of each website;
A word dividing step for dividing the sentence information into words;
Extracting a word that is the same as or similar to the predetermined keyword from the word group, and calculating the number of related information posts that calculates the number of text information including the word;
A related information posting rate calculating step for calculating a ratio of the number of related information postings to the number of text information collected from each Web site for each Web site;
An update frequency calculating step for calculating an update frequency of each Web site based on a reference date for analysis and the update date information;
A site activity analysis output step of outputting the related information posting rate for each website and the update frequency relative to each other;
An Internet site information analysis method comprising: In the Internet site information analysis method for accessing and analyzing the website information on the Internet, collecting the text information,
An information collecting step of collecting text information of the website every time a predetermined period elapses;
A word dividing step for dividing the sentence information into words;
A survey target information extracting step for extracting sentence information including the same or similar word as the survey target keyword from the word group;
A co-occurrence keyword extraction step for extracting a co-occurrence keyword that is included in the words constituting the survey target information and is a keyword different from the keyword;
A co-occurrence keyword score calculation step for calculating a score for each co-occurrence keyword based on the frequency of occurrence of the co-occurrence keyword in the survey target information;
A sorting step of rearranging the co-occurrence keywords in the order of the co-occurrence keyword scores to create a co-occurrence information list;
A co-occurrence information analysis output step for outputting the co-occurrence information list obtained for each predetermined period in a time series; and
An Internet site information analysis method comprising: In an Internet site information analysis apparatus that is configured by a computer system, accesses a Web site existing on the Internet, collects text information, and performs analysis,
Information collecting means for collecting text information of the website;
Word dividing means for dividing the sentence information into words;
Reputation information extraction means for extracting reputation information consisting of a set of an evaluation object, an evaluation axis representing the nature of the evaluation object, and an evaluation expression representing evaluation for the evaluation axis, from the word group;
Classification means for each evaluation axis for classifying the reputation information for each evaluation axis;
An evaluation expression dictionary group database in which a plurality of evaluation expression dictionaries including a set of the evaluation expression and a score indicating the degree are provided for each evaluation axis;
An evaluation expression dictionary corresponding to the evaluation axis to be analyzed is extracted from the evaluation expression dictionary group database, and a score calculation means for calculating a score for each piece of reputation information based on the evaluation expression dictionary;
Reputation analysis output means for comparing and outputting scores of arbitrary different evaluation axes for each sentence information;
An Internet site information analyzing apparatus comprising:   The reputation analysis output means outputs a two-dimensional graph based on a score of any two different evaluation axes or a three-dimensional graph based on a score of any three different evaluation axes. 4. Internet site information analysis apparatus according to 4. In an Internet site information analysis apparatus that is configured by a computer system, accesses a Web site existing on the Internet, collects text information, and performs analysis,
Information collecting means for collecting text information of the website;
Word dividing means for dividing the sentence information into words;
Extracting a word that is the same as or similar to a predetermined keyword from the word group, and calculating the number of related information posts that calculates the number of text information including the word;
Related information posting rate calculating means for calculating the ratio of the number of related information postings to the number of text information collected from each Web site for each Web site;
Update date information collecting means for collecting the update date of each website;
Update frequency calculation means for calculating the update frequency of each website based on the reference date for analysis and the update date information;
Site activity analysis output means for relatively comparing and outputting the related information posting rate for each Web site and the update frequency;
Internet site information analysis apparatus characterized by comprising   7. The Internet site information analysis apparatus according to claim 6, wherein the activity analysis output means outputs the related information posting rate and the update frequency in a two-dimensional graph. In the Internet site information analysis device that accesses and analyzes the website information that exists on the Internet,
Information collecting means for collecting text information of the website every time a predetermined period elapses;
Word dividing means for dividing the sentence information into words;
Investigation target information extracting means for extracting sentence information including a word that is the same as or similar to the keyword to be investigated from the word group;
A co-occurrence keyword extracting means for extracting a co-occurrence keyword which is included in the words constituting the investigation target information and is other words excluding the same or similar words as the investigation target keyword;
Co-occurrence keyword score calculation means for calculating a score for each co-occurrence keyword based on the frequency of appearance of the co-occurrence keyword in the survey target information;
Sort means for creating a co-occurrence information list by rearranging the co-occurrence keywords in the order of the co-occurrence keyword scores;
A co-occurrence information analysis output means for outputting the co-occurrence information list obtained for each predetermined period in time series; and
An Internet site information analyzing apparatus comprising:

Images