JP2013168177A - Information provision program, information provision apparatus, and provision method of retrieval service - Google Patents

Abstract

PROBLEM TO BE SOLVED: To improve efficiency of retrieval activity by a user in a retrieval system.SOLUTION: Articles published in blogs are extracted from a set of the blogs obtained by giving a retrieval query, and the extracted articles are classified into a plurality of clusters according to published contents. Then, conformity degree of the respective blogs to the retrieval query is calculated on the basis of correlation of the blogs and the clusters, and retrieved result relating to the blogs conforming to the retrieval query is presented on the basis of the calculated result. Thereby, since the respective blogs are evaluated considering the correlation of the blogs and topics on the basis of the published contents, appropriate scoring can be performed even for the blog having little link relation with other blogs.

Description

  The present invention relates to an information providing method, an information providing apparatus, an information providing program, and a recording medium in which the program is recorded in a computer. The information providing method presents a page group obtained by giving a search query received from a user as a search result.

  2. Description of the Related Art Conventionally, in the field of information search service, a technique for allowing a user to efficiently perform a search activity has been required. In response to this request, for example, when a search result is presented, a technique for calculating the score of each page based on the link relationship between pages and ranking the search result based on the score is known. Yes.

  Specifically, ranking of pages related to a search query is performed in consideration of factors such as links on a page, link anchors, and popularity of a page linked to another page. In this case, the relationship that “a page linked many from other pages is important” and “a page linked from important pages is important” is established.

  Patent Document 1 below describes a technique for performing page importance ranking for web search using a hierarchical structure and link relationship of the World Wide Web as a ranking method based on link relationship. Specifically, link graph analysis is performed on the aggregated link relationships to determine the importance of each node. The importance of each node propagates to the page associated with that node. Then, for each page, the page importance ranking is calculated using the importance of the page and the importance of the node associated with the page.

  Further, in Patent Document 2 below, as a search technique using a topic related to a search query, a topic is extracted from a set of topics for creating an index for web search or for web search, and search is performed based on the topic. Techniques for obtaining results are described. Specifically, an index used for searching a data group is generated from a topic map included in the information network, and index data of data that meets the search condition is displayed as a search result.

JP 2006-127529 A JP 2006-215753 A

  However, according to the prior art described in Patent Documents 1 and 2 described above, appropriate ranking can be performed for pages that have little or no link relationship with other pages. There is a problem that you can not. For example, a page that has just been registered on the network is ranked lower than a page with many links because it is less recognized by other users and is less likely to be linked from other pages. End up.

  In addition, ranking based on the link relationship is not a method that takes into account the search query and the posted content, so there may be a large divergence between the importance of the page and the link relationship. For this reason, because it does not have a link relationship, pages that have a strong relationship with the search query or pages on which information requested by the searcher is ranked lower, resulting in the user's work time and the search activity There is a problem that the work burden increases.

  In order to solve the above-described problems caused by the prior art, the present invention optimizes the ranking of search results and increases the efficiency of user search activities by performing appropriate scoring in consideration of the content of the page. It is an object to provide an information providing method, an information providing apparatus, an information providing program, and a recording medium in which the program is recorded in a computer.

  According to one aspect of the present invention, each set name of a plurality of sets formed in accordance with each posting content of a plurality of web pages searched according to a search query is displayed, and the plurality of displayed sets An information providing program and an information providing apparatus for displaying title information of web pages belonging to a set of selected set names are proposed.

  According to another aspect of the present invention, there is provided a search service executed by a server that outputs a search result to a client in response to a search request from the client, the search query transmitted from the client Providing a set name of each of a plurality of sets formed in accordance with each posting content of the plurality of web pages searched according to the client, of the set names of the plurality of sets by the client When any set name is selected, a search service providing method for providing title information of web pages belonging to the selected set of set names to the client is proposed.

  According to the information providing method, the information providing apparatus, the information providing program, and the recording medium in which the program is recorded on the computer, the ranking of the search result is optimized by performing appropriate scoring in consideration of the content of the page. Thus, it is possible to improve the efficiency of user search activities.

  Exemplary embodiments of an information providing method, an information providing apparatus, an information providing program, and a recording medium recording the program on a computer will be described below in detail with reference to the accompanying drawings.

(System configuration of search system)
First, the system configuration of the search system 100 according to the present embodiment will be described. FIG. 1 is a system configuration diagram of a search system. In FIG. 1, a search system 100 includes an information providing apparatus 101 capable of communicating via a network 110 such as the Internet, a LAN, and a WAN, a database server 102, and client terminals 103-1 to 103-n. .

  The information providing apparatus 101 is a computer apparatus having a function of searching for a page that matches a search query received from the client terminals 103-1 to 103-n from a group of pages disclosed on the network 110. Specifically, the search query received from the client terminals 103-1 to 103-n is transmitted to the database server 102 to search for a page that matches the search query.

  Here, the page may be, for example, a website published on the Internet, or may be a group of files in a folder on a limited network 110 such as a LAN. As a specific example of the website, there is a weblog (hereinafter referred to as “blog”) that can identify the creator (author). More specifically, a page is electronic information composed of text data, HTML layout information, images embedded in a document, audio, and moving images.

  Further, the page includes at least one publication information. Here, the posted information may be, for example, individual web pages constituting the website, or may be individual files in the above-described folder. Specifically, when the page is a blog, information (entries) for one time newly registered each time it is updated may be used as the posting information. Further, information in units of frames divided from the contents displayed on one screen may be used as posted information.

In addition, the information providing apparatus 101 calculates the degree of fitness for the search query for each searched page. Furthermore, page information representing the search result is generated based on the degree of matching of each page, and the page information is presented to the client terminals 103-1 to 103-n. This information providing apparatus 101 is applied to a search engine of an information search service, for example.

  The database server 102 is a computer device that includes a page DB (database) (not shown) and has a function of managing a group of pages on the network 110. The database server 102 searches the page DB for a page that matches the search query received from the information providing apparatus 101, and transmits the search result to the information providing apparatus 101.

  The client terminals 103-1 to 103-n are computer devices having a function of transmitting a search query input by a user operation to the information providing device 101. Further, as a result of transmitting the search query, the client terminals 103-1 to 103-n receive the search result that matches the search query from the information providing apparatus 101 and display the search result on the display screen.

(Hardware configuration of computer device)
Next, the hardware configuration of the information providing apparatus 101, the database server 102, and the client terminals 103-1 to 103-n (herein, simply referred to as “computer apparatus”) illustrated in FIG. 1 will be described. FIG. 2 is an explanatory diagram illustrating a hardware configuration of the computer apparatus.

  In FIG. 2, the computer device comprises a computer main body 210, an input device 220, and an output device 230, and can be connected to a network 110 such as a LAN, WAN, or the Internet via a router or a modem (not shown). It is.

  The computer main body 210 has a CPU, a memory, and an interface. The CPU controls the entire computer device. The memory is composed of ROM, RAM, HD, optical disk 211, and flash memory. The memory is used as a work area for the CPU.

  Various programs are stored in the memory, and loaded according to instructions from the CPU. Data read / write of the HD and the optical disk 211 is controlled by a disk drive. The optical disk 211 and the flash memory are detachable from the computer main body 210. The interface controls input from the input device 220, output to the output device 230, and transmission / reception with respect to the network 110.

  The input device 220 includes a keyboard 221, a mouse 222, a scanner 223, and the like. The keyboard 221 includes keys for inputting characters, numbers, various instructions, and the like, and inputs data. Further, it may be a touch panel type. The mouse 222 performs cursor movement, range selection, window movement, size change, and the like. The scanner 223 optically reads an image. The read image is captured as image data and stored in a memory in the computer main body 210. Note that the scanner 223 may have an OCR function.

  Examples of the output device 230 include a display 231, a speaker 232, and a printer 233. The display 231 displays data such as a document, an image, and function information as well as a cursor, an icon, or a tool box. The speaker 232 outputs sounds such as sound effects and reading sounds. The printer 233 prints image data and document data.

(Outline of this embodiment)
Next, an outline of the present embodiment will be described. FIG. 3 is an explanatory diagram showing an outline of the present embodiment. In FIG. 3, first, the client terminal 103-1 to 103-n (hereinafter simply referred to as “client terminal 103”) is input by the user operating the input device 220 such as the keyboard 221 or the mouse 222. A search query is transmitted to the information providing apparatus 101.

  Thereafter, the information providing apparatus 101 searches for a page that matches the search query received from the client terminal 103. As a result, the search result table 400 shown in FIG. 4, the article link table 500 shown in FIG. 5, and the blog link table 600 shown in FIG. 6 are created. In the present embodiment, a blog on which at least one article (published information) is posted will be described as an example.

  Next, using the search result table 400 and the article link table 500, language analysis of articles posted on each blog is executed for each article. Specifically, for example, language analysis such as morphological analysis and dependency analysis is executed for each article. As a result, an analysis result table 700 shown in FIG. 7 is created.

  Then, using the analysis result table 700, each article is clustered into a plurality of clusters corresponding to the contents of the publication. As a result, the article cluster table 1000 shown in FIG. 10 and the blog cluster table 1100 shown in FIG. 11 are created.

  Thereafter, a network model 1200 (see FIG. 12) is created using the blog link table 600, the article cluster table 1000, and the blog cluster table 1100. As a result, a network table 1300 shown in FIG. 13 is created.

  Next, the score of each node in the network model 1200 is calculated using the network table 1300. As a result, a score table 1400 shown in FIG. 14 is created. Then, the search results are ranked using the score table 1400.

  Finally, HTML information related to the blog group is generated according to the ranking, and the HTML information is presented to the client terminal 103 as a search result. As a result, the search result is displayed on the display 231 of the client terminal 103.

  In this embodiment, scoring is performed in consideration of the content of articles posted on a page (blog) that matches a search query regardless of whether or not there is a link relationship with other pages. As a result, it is possible to appropriately evaluate the fitness level of pages that have little or no link relationship with other pages, and to optimize the ranking.

(Memory contents of search result table)
Next, a search result table used for the information providing apparatus 101 will be described. This search result table is a table of search results obtained by giving a search query to the search system 100. Specifically, it is a table that represents a blog and an article posted on the blog in association with each other.

  FIG. 4 is an explanatory diagram showing the stored contents of the search result table. In FIG. 4, the search result table 400 stores a blog ID and an article ID for each article posted on a blog that matches the search query. The blog ID is an identifier for identifying a blog. The article ID is an identifier for identifying an article. Here, taking blog B55 as an example, articles P85 and P71 are posted on blog B55.

  These blog IDs and article IDs are expressed by, for example, URLs (Uniform Resource Locators) that indicate locations on the network 110 where blogs and articles exist. The search result table 400 realizes its function by the storage unit such as ROM, RAM, HD described in FIG.

(Stored contents of article link table)
Next, an article link table used for the information providing apparatus 101 will be described. This article link table is a table of link relationships obtained by detecting hyperlinks embedded in articles. More specifically, the table is a table that represents the articles of the link source (from) and the articles of the link destination (to) in association with each other.

  FIG. 5 is an explanatory diagram showing the contents stored in the article link table. In FIG. 5, the article link table 500 stores the number of hyperlinks embedded in the article for each article. Specifically, for each article, the number of hyperlinks linked from the article (from) to another article (to) is stored.

  Here, taking article P53 as an example, two hyperlinks that link to article P85 are embedded in article P53. Note that the article link table 500 realizes its function by the storage unit such as ROM, RAM, HD described in FIG.

(Contents stored in the blog link table)
Next, a blog link table used for the information providing apparatus 101 will be described. This blog link table is a table showing the correlation between blogs obtained using the search result table 400 shown in FIG. 4 and the article link table 500 shown in FIG.

  FIG. 6 is an explanatory diagram showing the stored contents of the blog link table. In FIG. 6, a blog link table 600 stores a numerical value representing a correlation between each blog and another blog having a link relationship with the blog. This numerical value corresponds to the transition probability (details will be described later) of transition from one blog to the other blog.

  Here, taking blog B25 as an example, it has a link relationship with blog B23 and blog B55, and the numerical value representing the correlation is “0.5”. Also, this numerical value is normalized so that the total number of rows in the blog link table 600 is “1”. The blog link table 600 realizes its function by the storage unit such as ROM, RAM, HD described in FIG.

(Storage contents of analysis result table)
Next, an analysis result table used for the information providing apparatus 101 will be described. This analysis result table is a table of analysis results obtained by performing language analysis (for example, morphological analysis, dependency analysis) on each article.

  FIG. 7 is an explanatory diagram showing the storage contents of the analysis result table. In FIG. 7, the analysis result table 700 stores an article ID and an analysis result for each article. Here, taking the article P53 as an example, the analysis result R1 is an analysis result of the language analysis performed on the article P53. A specific example of the analysis result will be described later with reference to FIG. The analysis result table 700 realizes its function by the storage unit such as ROM, RAM, HD described in FIG.

(Functional configuration of information providing device)
Next, a functional configuration of the information providing apparatus 101 will be described. FIG. 8 is a block diagram illustrating a functional configuration of the information providing apparatus. 8, the information providing apparatus 101 includes a receiving unit 801, a searching unit 802, an extracting unit 803, a classifying unit 804, a calculating unit 805, a creating unit 806, a presenting unit 807, a determining unit 808, A selection unit 809.

  Each of the functions 801 to 809 can be realized by causing the CPU to execute a program related to the functions 801 to 809 stored in the storage unit of the information providing apparatus 101 or by using an input / output I / F. it can. Output data from each function 801 to 809 is held in the storage unit. In addition, the connection destination function indicated by the arrow in FIG. 8 reads output data from the connection source function from the storage unit and causes the CPU to execute a program related to the function.

  First, the receiving unit 801 has a function of receiving a search query input. The search query is, for example, a character string that represents an inquiry to the search system 100 (see FIG. 1). Specifically, for example, the accepting unit 801 accepts an input of a search query transmitted from the client terminal 103 (see FIG. 1).

  The search unit 802 has a function of searching for a page that matches the search query received by the receiving unit 801 from a group of pages published on the network 110. A page is a website such as a blog published on the Internet, for example. Specifically, for example, by transmitting the search query received by the receiving unit 801 to the database server 102, a search result that matches the search query can be obtained.

  The extraction unit 803 has a function of extracting publication information posted on the page selected from a set of pages including at least one or more publication information searched by the search unit 802. Here, the posted information is an article, an image, a sound, a moving image, or the like constituting the page. More specifically, for example, individual web pages constituting a website. In the present embodiment, description will be given by taking an article (for example, information newly registered once every time a blog is updated) as an example of posted information.

  Specifically, for example, the extraction unit 803 extracts articles posted on each blog from the search result table 400 illustrated in FIG. Note that the extraction unit 803 may extract the publication information posted on all the searched pages, and any page (for example, 100 blogs) among the searched pages (for example, 100 blogs). It is good also as extracting the publication information published on 50 blogs selected at random).

  The classification unit 804 has a function of classifying the posted information (articles) extracted by the extracting unit 803 into a plurality of clusters corresponding to the posted contents. Specifically, for example, the articles may be classified into a plurality of clusters based on the appearance frequency of each word appearing in the article extracted by the extraction unit 803 and the similarity between words.

  More specifically, for example, for each article extracted by the extraction unit 803, morphological analysis (division into morphemes and part-of-speech assignment) and dependency analysis (combination of phrases from morphemes and relationships between phrases) Clustering is performed using the analysis result obtained by executing (identification of receiving relationship).

  Here, the analysis result of the language analysis will be described. FIG. 9 is an explanatory diagram illustrating an example of the analysis result. In FIG. 9, the analysis result 900 shows the analysis result of the language analysis for the article “Perform language analysis for the next clustering for all documents ...”. Specifically, the analysis result 900 includes information on the phrase ID, the modification-destination phrase ID, and the part of speech for each phrase (word) divided from the analysis target article.

  For example, reference numeral 901 is a phrase ID for identifying “all”. Reference numeral 902 is the clause ID of the clause “NO” to which the clause “ALL” is modified. In this way, by performing linguistic analysis for each article, the article is divided into morphemes that are the smallest semantic units constituting each sentence in the article, and each clause in the sentence relates to which clause (which clause To modify).

  Hereinafter, a specific processing procedure of classification processing (clustering) by the classification unit 804 will be described by taking the analysis result table 700 illustrated in FIG. 7 as an example. First, the classification unit 804 refers to the analysis result table 700 and counts the number of appearances of words appearing in the articles P53, P85, P40, and P71. And the similarity between articles is calculated by calculating | requiring a cosine distance using a word vector.

  Next, a cluster is created according to a preset number of clusters, and a word vector for each cluster is created. Specifically, the word vectors of each cluster are created by adding the word vectors of the documents in the cluster. The number of clusters can be arbitrarily set by the user operating the input device 220 such as the keyboard 221 and the mouse 222 shown in FIG.

  Thereafter, the cosine distance between the cluster word vector and the article word vector is obtained. Then, the obtained cosine distance is set as the affiliation probability for each article cluster. Note that the language analysis of the posted information extracted by the extraction unit 803 is performed in the information providing apparatus 101, but the present invention is not limited to this. For example, the language analysis of the posted information may be executed using an external computer device, and the execution result may be acquired as the analysis result of the language analysis.

  Here, a specific example of the article cluster table representing the classification results classified into a plurality of clusters by the classification unit 804 will be described. FIG. 10 is an explanatory diagram showing the contents stored in the article cluster table. In FIG. 10, the article cluster table 1000 stores the affiliation probabilities belonging to each of the clusters C1 to C4 by associating the article ID with the cluster ID for each article. The cluster ID is an identifier for identifying a cluster.

  The affiliation probability is a numerical value that represents the strength of correlation of each article with respect to the clusters C1 to C4. Taking the article P53 as an example, the membership probability to the cluster C1 is “0.1”, the membership probability to the cluster C2 is “0.5”, and the membership probability to the cluster C3 is “0.2”. The belonging probability to the cluster C4 is “0.0”. Note that the article cluster table 1000 realizes its function by the storage unit such as the ROM, RAM, and HD described in FIG.

  Returning to the description of FIG. 8, the calculation unit 805 calculates, for each page, the suitability of the page with respect to the search query based on the correlation between the page searched by the search unit 802 and the cluster classified by the classification unit 804. It has a function. Specifically, the fitness of each page can be calculated using a network model that is a directed graph with nodes (for example, blogs) and clusters created by the creation unit 806 described later as nodes.

  The creation unit 806 has a function of creating a network model in a directed graph with pages and clusters as nodes. Specifically, for example, the creation unit 806 first creates a blog cluster table representing the affiliation probability of a blog (page) with respect to a cluster using the article cluster table 1000 shown in FIG.

  More specifically, for each blog, the affiliation probabilities for all articles posted on the blog are added, and the addition result is set as the affiliation probability for the blog cluster. Here, taking the affiliation probability of the blog B55 with respect to the cluster C2 as an example, referring to the article cluster table 1000, the affiliation probability “0.8” with respect to the cluster C2 of the article P85 and the affiliation probability with respect to the cluster C2 of the article P71 “ The membership probability “1.2” is obtained by adding “0.4”.

  Here, a specific example of the blog cluster table created by the creation unit 806 will be described. FIG. 11 is an explanatory diagram showing the storage contents of the blog cluster table. In FIG. 11, the blog cluster table 1100 stores transition probabilities of transition to each of the clusters C1 to C4 in association with the blog ID and the cluster ID for each blog.

  Taking blog B23 as an example, the transition probability to cluster C1 is “0.1”, the transition probability to cluster C2 is “0.5”, the transition probability to cluster C3 is “0.2”, The transition probability to the cluster C4 is “0.0”. The blog cluster table 1100 realizes its function by the storage unit such as ROM, RAM, HD described in FIG.

  Thereafter, the creation unit 806 uses the blog cluster table 1100 and the blog link table 600 shown in FIG. 6 to create a network model in which the blogs B23, B55, and B25 and the clusters C1 to C4 are used as nodes as a directed graph. Create

  FIG. 12 is a directed graph showing an example of a network model. In FIG. 12, a network model 1200 is a directed graph in which blogs B23, B25, and B55 and clusters C1 to C4 are noded. Further, transition probabilities between the nodes are given to the edges connecting the nodes. In the drawing, a part of the transition probability between nodes is extracted and displayed.

  In the network model 1200, a double-headed arrow indicated by a solid line represents a transition between a node representing a blog and a node representing a cluster. The two-way arrow has a transition probability from a node representing a blog to a node representing a cluster (underlined in FIG. 12), and a transition probability from a node representing the cluster to a node representing a blog (double underlined in FIG. 12). ) And are granted.

  Furthermore, the affiliation probability (the angle brackets in FIG. 12) that the blog managed in the blog cluster table 1100 belongs to the cluster is given to this bidirectional arrow. Further, in the network model 1200, arrows indicated by dotted lines represent transitions between nodes representing a blog. This arrow is given a transition probability between nodes representing the blog managed in the blog link table 600.

  The transition probability between a node representing a blog and a node representing a cluster can be obtained using the affiliation probability with respect to a cluster of articles posted on the blog. More specifically, for example, the transition probability X (Bi−Cj) for transition from the blog Bi to the cluster Cj can be obtained using the following equation (1). However, Bi is a blog ID, Cj is a cluster ID, k is a natural number, and n is the number of clusters (here, n = 4).

  Here, taking as an example the transition probability X (B23) of transition from the node representing the blog B23 to the node representing the cluster C1, “X (B23−C1) = (0.1 / 0.8)”. This is because the affiliation probability “0.1 + 0...” Is obtained by adding the affiliation probability “0.1” of the blog B23 managed in the blog cluster table 1100 to the clusters C1 to C4 of the blog B23. It is a value divided by “5 + 0.2 + 0.0 = 0.8”.

  Further, the transition probability X (Cj−Bi) of transition from the node representing the cluster Cj to the node representing the web Bi can be obtained using the following equation (2). However, m is the number of blogs (here, m = 3).

  Here, taking as an example the transition probability X (C1-B23) of transition from the node representing the cluster C1 to the node representing the web B23, “X (C1-B23) = (0.1 / 0.3)”. Become. This is because the affiliation probability “0...” Obtained by adding the affiliation probability “0.1” of the blog B23 managed in the blog cluster table 1100 to the cluster C1 and the affiliation probability of each blog B23, B55, B25 to the cluster C1. 1 + 0.2 + 0.0 = 0.3 ”.

  The network model 1200 shown in FIG. 12 is held in the storage unit as a network table described below. FIG. 13 is an explanatory diagram showing the contents stored in the network table. In FIG. 13, the network table 1300 stores the transition probabilities between the nodes in the network model 1200 (see FIG. 12) in association with the cluster ID and the blog ID.

  Here, the transition probabilities between the nodes representing the clusters are all “0”. That is, there is no transition between nodes representing a cluster. Further, the transition probability between the node representing the blog and the node representing the cluster is the transition probability obtained using the above formula (1) or (2). Further, the transition probability between nodes representing a blog refers to the blog link table 600 and is normalized so that the total number of rows is “1”.

  For example, the calculation unit 805 calculates the fitness of the blog for each blog based on the storage contents of the network table 1300. Specifically, the fitness of each blog can be calculated by scoring each node by a random walk method using transition probabilities assigned to edges connecting nodes in the network model 1200. it can.

  A random walk is a motion in which the “probability of what appears next” is determined irregularly. Here, the concept of random walk is applied to scoring of each node. Here, an outline of processing for scoring each node by a random method will be described.

  First, an end probability Pt (for example, “0.2”) at which the random walk ends is set in advance. Then, using the transition probability X () between nodes and the end probability Pt, the time from when one node is arbitrarily selected from the node group in the network model 1200 to the end of the transition between the nodes. In addition, the number of transitions to each node is counted for each node.

  Then, a series of processes from arbitrarily selecting one node until the end of transition between the nodes is repeated a predetermined number of times N (for example, “10000”). Finally, as a result of repeating the specified number of times N, the number of transitions to each node is set as the fitness of each node with respect to the search query. The end probability Pt and the prescribed number N described above can be arbitrarily set by the user operating the input device 220.

  Here, by forming a directed graph using blogs and clusters as nodes, it is possible to form a virtual network (network model 1200) to which a transition probability representing the correlation between the blog and the clusters is given. As a result, the random walk method can be applied to scoring each node.

  In scoring by the random walk method, the higher the number of transitions to the node, the higher the score. This is based on the relationship that as the number of transitions increases, the possibility (expected value) of visiting a node during a random walk increases. Here, by using this relationship, the clusters or blogs that represent nodes with a high number of transitions are treated as clusters or blogs that are more likely to be viewed by searchers (actually only pages are viewed) , Increase the relevance to the search query.

  Here, a specific example of a score table representing the fitness (score) of each node calculated by the calculation unit 805 will be described. FIG. 14 is an explanatory diagram showing the contents stored in the score table. In FIG. 14, the score table 1400 stores a score representing the degree of fitness for the search query for each node in the network model 1200.

  Here, taking a node representing the cluster C1 as an example, the degree of matching with the search query is “2523”. Taking a node representing the blog B23 as an example, the fitness for the search query is “524”.

  The presentation unit 807 has a function of presenting page information related to a page group based on the calculation result calculated by the calculation unit 805. Specifically, page information related to the page group is generated based on the calculation result calculated by the calculation unit 805, and the page information is transmitted to the client terminal 103 (search query transmission source).

  More specifically, for example, with reference to the score table 1400 illustrated in FIG. 14, an index (index information) in which the blogs B23, B55, and B25 are rearranged in descending order of suitability for the search query is generated as a search result. It is good as well.

  At this time, since there is no website corresponding to the clusters C1 to C4, an index considering only the blogs B23, B55, and B25 is generated. That is, in the score table 1400, the blogs B23, B55, and B25 are rearranged by evaluating the effectiveness with respect to the search query from the fitness of the blogs B23, B55, and B25 excluding the fitness levels of the clusters C1 to C4.

  The determination unit 808 has a function of determining, for each cluster, a word (topic) that characterizes the cluster, based on the appearance frequency of words appearing in the posting information classified into a plurality of clusters by the classification unit 804. Specifically, for each cluster, a word having a high appearance frequency that appears in the classified posted information (article) is determined as a topic that represents the feature of the cluster.

  In addition, the determination unit 808 determines a word characterizing the cluster based on a value obtained by dividing the number of appearances of the word appearing in the posting information by the number of appearances of the word appearing in all the posting information posted on the page group. It may be determined. According to this, the cluster is characterized from other words excluding words such as particles (for example, “O”) and symbols (for example, “.”) That tend to appear more frequently regardless of the content. A word can be determined.

  Further, the determination unit 808 may determine a word characterizing the cluster based on the appearance frequency of a word with a specific part of speech that appears in the posted information. According to this, the words that characterize the cluster can be limited to specific parts of speech such as nouns and verbs.

  Here, a specific example of the feature word table representing the appearance frequency of words appearing in the posted information for each cluster will be described. FIG. 15 is an explanatory diagram showing the stored contents of the feature word table. In FIG. 15, the feature word table 1500 stores numerical values representing the appearance frequency of words appearing in articles classified into each cluster for each of the clusters C1 to C4.

  Specifically, the value obtained by dividing the number of appearances of the word by the number of appearances of the word appearing in all articles (P53, P85, P40, P71) published in the blog group (B23, B25, B55) is It is remembered. For example, the determination unit 808 refers to the feature word table 1500 and determines the word characterizing the cluster C1 as “sound quality”.

  Returning to the description of FIG. 8, the presentation unit 807 has a function of presenting the determination result determined by the determination unit 808 as a topic related to the search query. Specifically, for example, a topic may be presented together with page information related to a page that matches a search query.

  At this time, for example, each page may be presented in association with a topic that characterizes a cluster having the maximum affiliation probability to which the page belongs. Specifically, for example, when the page information related to the blog B23 is presented, the blog cluster table 1100 is referred to, and a word characterizing the cluster C2 having the maximum affiliation probability is presented as a topic. In addition, referring to the score table 1400, the topics of the clusters C1 to C4 may be rearranged and presented in descending order of the degree of fitness for the search query.

  Here, an example of a screen displayed on the display 231 of the client terminal 103 as a result of the page information presented by the presentation unit 807 will be described. FIG. 16 is an explanatory diagram (part 1) illustrating an example of a screen displayed on the display. In FIG. 16, a search result 1600 obtained by giving a search query “FJpod” is displayed on the display 231.

  Specifically, the page titles are displayed in descending order of suitability for the search query “FJpod”. Each page title is displayed in association with a topic that characterizes a cluster corresponding to the page title. Here, when the cursor C is moved and an arbitrary title is clicked, the web page of the title is displayed on the display 231.

  As another screen example, search results may be classified and presented for each topic. FIG. 17 is an explanatory diagram (part 2) illustrating an example of a screen displayed on the display. In FIG. 17, a search result 1700 obtained by giving a search query “FJpod” is displayed on the display 231.

  Specifically, a page title is displayed for each topic related to the search query “FJpod”. Here, when the cursor C is moved and the “extension” topic is clicked, the page title of a website that publishes articles with a high occurrence frequency of “extension” is displayed.

  The selection unit 809 has a function of accepting selection of an arbitrary topic from among a plurality of topics presented by the presentation unit 807. Specifically, as a result of displaying a plurality of topics on the display 231 of the client terminal 103, the topic selected by the user's operation input is accepted.

  In this case, in the fitness calculation process using the random walk method, a node representing a cluster characterized by the topic selected by the selection unit 809 is selected instead of arbitrarily selecting the first node to start the random walk. It is good as well. In this case, the calculation unit 805 counts the number of transitions made to each node between the time when the node representing the cluster characterized by the topic selected by the selection unit 809 is selected and the time when the transition between the nodes ends. By doing so, the degree of conformity of the page is calculated. Thereby, the score of the page (blog) on which the posting information (article) related to the selected topic is posted can be increased.

  Here, a search procedure in the client terminal 103 will be described. FIG. 18 is an explanatory diagram illustrating an example of a search procedure in the client terminal. In FIG. 18, first, when a user gives a search query “FJpod”, a search result 1810 is displayed on the display 231. Specifically, the search result 1810 displays a plurality of topics (design, sound quality, accessories, dealers) related to the search query “FJpod”.

  Next, in the search result 1810, the cursor C is moved and any button B1 to B4 is clicked to select an arbitrary topic. Here, it is assumed that the buttons B1 and B2 are clicked. Thereafter, when the enter button on the keyboard 221 is pressed, the search result 1820 is displayed on the display 231.

  Specifically, the search result 1820 displays page information obtained by increasing the degree of fitness of a page (blog) on which articles about “design” and “sound quality” are posted. In this way, by selecting an arbitrary topic from among a plurality of topics related to the search query, it is possible to present a search result that more reflects the searcher's will.

  Note that the functions of the receiving unit 801 and the searching unit 802 may be realized by an external computer device. In this case, the extraction unit 803 extracts the posted information posted on the page from the page group acquired from the external computer device.

(Various processing procedures of the information providing device)
Next, various processing procedures executed in the information providing apparatus 101 according to the present embodiment will be described.

(Blog link table creation processing procedure)
First, a creation processing procedure for creating the blog link table 600 shown in FIG. 6 will be described. The blog link table 600 is a table showing the correlation between the blogs B23, B55, and B40. Here, the creation unit 806 creates the blog link table 600 using the search result table 400 shown in FIG. 4 and the article link table 500 shown in FIG.

  FIG. 19 is a flowchart illustrating an example of a creation processing procedure for creating a blog link table. In the flowchart of FIG. 19, first, a blog link table having blog IDs in rows and columns is created, and each cell is initialized with “0” (step S1901).

  Thereafter, an arbitrary cell in which the transition probability is to be stored is selected from the blog link table (step S1902), and the blog ID of the transition source (from) is selected for the selected cell (hereinafter referred to as “selected cell”). It is determined whether or not the blog ID of the transition destination (to) is the same (step S1903).

  Here, when it is determined that the transition source (from) and the transition destination (to) are different (step S1903: No), the value of the corresponding article ID from the article link table 500 representing the correlation between articles. Is added to the selected cell (step S1904). Specifically, an article posted on the blog of the selected cell is identified with reference to the search result table 400, and the value of the identified article from the article link table 500 is read and added.

  If it is determined in step S1903 that the transition source (from) and the transition destination (to) are the same (step S1903: Yes), the process proceeds to step S1905.

  Next, it is determined whether or not there is an unselected cell not selected from the blog link table (step S1905). If there is an unselected cell (step S1905: Yes), the process returns to step S1902, and the blog A series of processing is repeated by selecting an unselected cell from the link table.

  On the other hand, if there is no unselected cell (step S1905: No), an arbitrary row is selected from the blog link table (step S1906) so that the total value of each cell becomes “1”. Normalization is performed (step S1907). Thereby, the value of each cell of the blog link table can be leveled.

  Next, it is determined whether or not there is an unselected row not selected from the blog link table (step S1908). If there is an unselected row (step S1908: Yes), the process returns to step S1906, and the blog An unselected row is selected from the link table and a series of processing is repeated.

  On the other hand, if there is no unselected row (step S1908: No), the blog link table 600 is output (saved) to the storage unit (step S1909), and the series of processing according to this flowchart ends.

  Thereby, the blog link table 600 representing the correlation between blogs can be created. Further, by normalizing the value of each cell for each row in the blog link table 600, the value of each cell can be leveled throughout the blog link table 600.

(Blog cluster table creation processing procedure)
Next, a creation processing procedure for creating the blog cluster table 1100 shown in FIG. 11 will be described. The blog cluster table 1100 is a table showing the correlation between the blogs B23, B55, and B25 and the clusters C1 to C4. Here, the creation unit 806 creates the blog cluster table 1100 using the search result table 400 shown in FIG. 4 and the article cluster table 1000 shown in FIG.

  FIG. 20 is a flowchart illustrating an example of a creation processing procedure for creating a blog cluster table. In the flowchart of FIG. 20, first, a blog cluster table having a blog ID in a column and a cluster ID in a row is created, and each cell is initialized with “0” (step S2001).

  Thereafter, an arbitrary row for storing the blog ID and the article ID is selected from the search result table 400 (step S2002), and the selected row from the article cluster table 1000 (hereinafter referred to as “selected row”). ) Is read (step S2003).

  Then, the read value is added to the row corresponding to the blog ID of the selected row of the blog cluster table (step S2004). Specifically, the probability of belonging to each cluster of the article specified from the article ID of the selected row is added as the transition probability to each cluster of the blog specified from the blog ID of the selected row. That is, the correlation between the blog and the cluster is expressed using the correlation between the article and the cluster.

  Next, it is determined whether or not there is an unselected row not selected from the search result table 400 (step S2005). If there is an unselected row (step S2005: Yes), the process returns to step S2002, An unselected row is selected from the search result table 400 and a series of processing is repeated.

  On the other hand, if there is no unselected row (step S2005: No), the blog cluster table is output (saved) to the storage unit (step S2006), and the series of processing according to this flowchart ends. Thereby, the blog cluster table 1100 representing the correlation between the blog and the cluster can be created.

(Network table creation processing procedure)
Next, a creation processing procedure for creating the network table 1300 shown in FIG. 13 will be described. The network table 1300 is a table that represents the correlation between the blogs B23, B55, and B25 and the clusters C1 to C4 as a directed graph (network model 1200 shown in FIG. 12). Here, the creation unit 806 creates the network table 1300 using the blog link table 600 shown in FIG. 6 and the blog cluster table 1100 shown in FIG.

  FIG. 21 is a flowchart illustrating an example of a creation processing procedure for creating a network table. In the flowchart of FIG. 21, first, a network table having cluster IDs and blog IDs in rows and columns is created, and each cell in which transition probabilities are to be stored is initialized with “0” (step S2101).

  Thereafter, an arbitrary cell in which the transition probability is to be stored is selected from the network table (step S2102), and the blog ID of the transition source (from) or the selected cell (hereinafter referred to as “selected cell”) is selected. It is determined whether the cluster ID is the same as the transition destination (to) blog ID or cluster ID (step S2103). If it is determined that the transition source (from) and the transition destination (to) are the same (step S2103: Yes), the process proceeds to step S2113.

  On the other hand, if it is determined that the transition source (from) and the transition destination (to) are different (step S2103: No), it is determined whether or not both of the selected cells are cluster IDs (step S2104). If both cluster IDs are determined (step S2104: YES), the process proceeds to step S2113.

  On the other hand, if both are not determined to be cluster IDs (step S2104: No), it is determined whether the selected cell is both blog IDs (step S2105). If both blog IDs are determined (step S2105: YES), the corresponding blog ID value is read from the blog link table 600 (step S2106), and the value is written in the selected cell (step S2107). ), The process proceeds to step S2113.

  On the other hand, if both blog IDs are not determined (step S2105: No), it is determined whether the selected cell is a blog ID from the cluster ID (step S2108). Here, when the blog ID is determined from the cluster ID (step S2108: Yes), the transition probability X (Cj−Bi) for transition from the cluster Cj to the blog Bi is calculated using the above equation (2) (step S2108: Yes). S2109). Then, the transition probability X (Cj-Bi) is written in the selected cell (step S2110), and the process proceeds to step S2113.

  On the other hand, when the blog ID is not determined from the cluster ID (step S2108: No), the transition probability X (Bi-Cj) for transition from the blog Bi to the cluster Cj is calculated using the above equation (1) (step S2108: No). S2111). Then, the transition probability X (Bi-Cj) is written in the selected cell (step S2112).

  As described above, the transition probability between the blog and the cluster is obtained using the affiliation probability with respect to the cluster of the article posted on the blog, so that the node representing the blog and the node representing the cluster in the network model 1200 are obtained. The link relationship can be formed.

  Next, it is determined whether there is an unselected cell that has not been selected from the network table (step S2113). If there is an unselected cell (step S2113: YES), the process returns to step S2102 to select an unselected cell from the network table and repeat a series of processes.

  On the other hand, if there is no unselected cell (step S2113: No), the network table is output (saved) to the storage unit (step S2115), and the series of processing according to this flowchart ends. Thereby, the network table 1300 showing the correlation between the nodes in the network model 1200 can be created.

(Fitness calculation processing procedure)
Next, a calculation processing procedure for calculating the fitness of a blog with respect to a search query will be described. Here, the calculation unit 805 calculates the fitness of each blog by applying a random walk method to the network model 1200 shown in FIG. 12 (network table 1300 shown in FIG. 13). However, the prescribed number of random walks is N, the number of trials is n, and the end probability is Pt.

  FIG. 22 is a flowchart illustrating an example of a calculation processing procedure for calculating the fitness level of a blog. In the flowchart of FIG. 22, first, a transition probability X (Cj−Bi) for transition from the cluster Cj to the web Bi is initialized with “0” (step S2201).

  A score table having a cluster ID representing a node and a blog ID in a column is created, each cell in which a score is to be stored is initialized with “0” (step S2201), and the number of trials n is initialized with “0”. (Step S2202).

  Thereafter, “1” is incremented to the number of trials n (step S2203), one node is randomly selected from the from column of the network table 1300 (step S2204), and the selected node ( Hereinafter, “1” is incremented in the cell of “selected node” (step S2205).

  Then, one element is randomly selected from the element group in which the end probability Pt is added to each element of the selected node string (step S2206). Thereafter, it is determined whether or not the end of the random walk is selected (step S2207). If the end is not selected (step S2207: No), the node corresponding to the selected element is selected (step S2208). ), The process returns to step S2205.

  That is, the random walk in the network model 1200 is continued until the end of the transition is selected, and each time a node is reached, the number of transitions that becomes the score (fitness) of that node is counted.

  On the other hand, when the end of the random walk is selected (step S2207: Yes), it is determined whether or not the number of trials n ≧ the prescribed number N (step S2209). Here, when the number of trials n <the prescribed number N (step S2209: No), the process returns to step S2203 to repeat a series of processes.

  On the other hand, when the number of trials n ≧ the prescribed number N (step S2209: Yes), the score table is output (saved) to the storage unit (step S2210), and the series of processing according to this flowchart ends.

  Accordingly, it is possible to create a score table 1400 that represents the degree of suitability of each blog with respect to the search query based on the number of times of transition to the node representing the blog. According to this, blogs that post articles on topics with high scores (fitness) have higher scores, and topics included in articles that are posted on blogs with higher scores have higher scores. .

(Feature word table creation process)
Next, a creation processing procedure for creating the feature word table 1500 shown in FIG. 15 will be described. The feature word table 1500 is a table that lists the appearance frequencies of words appearing in articles classified into the clusters C1 to C4. Here, the creation unit 806 creates the feature word table 1500 using the analysis result table 700 shown in FIG. 7 and the article cluster table 1000 shown in FIG.

  FIG. 23 is a flowchart illustrating an example of a creation processing procedure for creating a feature word table. In FIG. 23, first, a feature word table having a cluster ID in a column and words (all words appearing in a blog group) in a row is created, and each cell in which a numerical value is to be stored is initialized to “0” (step) S2301). Thereafter, an arbitrary cluster is selected from the article cluster table 1000 (step S2302).

  Next, an arbitrary article is selected from the column of the selected cluster (hereinafter referred to as “selected cluster”) (step S2303). Then, with reference to the analysis result table 700, an arbitrary word is selected from the analysis results of the selected article (hereinafter referred to as “selected article”) (step S2304). Thereafter, the affiliation probability of the selected article is added to the corresponding cell (column: selected cluster, row: selected word) of the feature word table (step S2305).

  Next, it is determined whether or not there is an unselected word that is not selected from the analysis result of the selected article (step S2306). If there is an unselected word (step S2306: Yes), the process returns to step S2304. Then, an unselected word is selected from the analysis results and a series of processing is repeated.

  On the other hand, if there is no unselected word (step S2306: No), it is determined whether there is an unselected article that has not been selected from the selected cluster column (step S2307). If there is an unselected article (step S2307: YES), the process returns to step S2303, an unselected article is selected from the selected cluster column, and a series of processes is repeated.

  On the other hand, if there is no unselected article (step S2307: No), it is determined whether there is an unselected cluster that has not been selected from the article cluster table 1000 (step S2308). If there is an unselected cluster (step S2308: Yes), the process returns to step S2302, and an unselected cluster is selected from the article cluster table 1000, and a series of processing is repeated.

  On the other hand, when there is no unselected cluster (step S2308: No), a weighting process for weighting the value of each cell in the feature word table is executed (step S2309), and finally the feature word table is output to the storage unit. (Save) (step S2310), and a series of processing according to this flowchart is terminated. Thereby, it is possible to create the feature word table 1500 that lists the appearance frequencies of words appearing in articles classified into the clusters C1 to C4.

  Next, a specific processing procedure of the weighting process in step S2309 in FIG. 23 will be described. This weighting process weights the values of words that frequently appear in a specific article by leveling the values of words that frequently appear in all articles. FIG. 24 is a flowchart illustrating an example of a specific processing procedure of weighting processing.

  In the flowchart of FIG. 24, first, the number of appearances of all words is counted for each word based on the stored contents of the analysis result table 700 (step S2401). Thereafter, an arbitrary word is selected from the feature word table (step S2402).

  Then, the value of each cell of the selected word string is divided by the number of appearances of the word counted in step S2401 (step S2403). Next, it is determined whether there is an unselected word that has not been selected from the feature word table (step S2404).

  If there is an unselected word (step S2404: YES), the process returns to step S2402, selects an unselected word from the feature word table, and repeats a series of processes. On the other hand, when there is no unselected word (step S2404: No), the process proceeds to step S2310 shown in FIG.

  As a result, cell values corresponding to words such as particles and symbols that appear frequently regardless of the content of the article are reduced, and as a result, cell values corresponding to words that frequently appear in a specific article Can be weighted.

(Search result presentation processing procedure)
FIG. 25 is a flowchart illustrating an example of a search result presentation processing procedure. However, the prescribed number of blogs presented as search results is K. In the flowchart of FIG. 25, first, k representing the number of blogs is initialized to “0” (step S2501). Thereafter, the blog ID having the highest score is identified with reference to the score table 1400 (step S2502), and “1” is incremented to the blog count k (step S2503).

  Next, the identified blog ID and the article ID associated with the blog ID are extracted from the search result table 400 (step S2504). In addition, referring to the blog cluster table 1100, the cluster ID having the highest affiliation probability is identified from the identified blog ID rows (step S2505). Thereafter, referring to the feature word table 1500, the word having the maximum value is extracted from the identified cluster ID line (step S2506).

  Next, it is determined whether the number of blogs k ≧ the specified number K or whether all the blog IDs in the score table 1400 have been specified (step S2507). If it is determined that the number of blogs k <the specified number K (step S2507: No), the process returns to step S2502, and the blog ID having the highest score and the unspecified is specified, and a series of processes is repeated.

  On the other hand, if it is determined that the number of blogs k ≧ the specified number K or all the blog IDs in the score table 1400 have been specified (step S2507: Yes), the blog IDs, article IDs, and the IDs extracted in steps S2504 and S2506 Page information is generated using words (step S2508).

  Finally, the generated page information is presented to the client terminal 103 (step S2509), and a series of processing according to this flowchart is terminated. According to this, it is possible to present a search result based on the optimized ranking to the searcher by performing appropriate scoring in consideration of the content of the blog.

  As described above, according to the present embodiment, using the correlation between pages and clusters (topics) based on the content of each page (blog) (the content of articles posted on the blog), Each page can be scored. Thereby, for example, appropriate scoring can be performed even for a blog that has little or no link relationship with other blogs. As a result, the relationships “important blogs that deal with important topics” and “important topics handled by important blogs” are established, and rankings are performed in consideration of the content of the blog.

  Further, when scoring each node by the random walk method, the influence of the link relationship can be reduced by using the normalized transition probability between pages (between blogs). Thereby, for example, it is possible to prevent a blog having many link relationships with other blogs from being ranked higher regardless of the posted content.

  Further, by causing the searcher to select an arbitrary topic from among a plurality of topics related to the search query, it is possible to increase the score of a page on which posted information related to a specific topic is posted. As a result, for example, a blog on which an article related to a topic that the searcher is interested in is easily ranked higher, and a search result that reflects the searcher's will can be presented.

  As described above, according to the information providing method, the information providing apparatus, the information providing program, and the recording medium in which the program is recorded on the computer, the search result can be obtained by performing appropriate scoring in consideration of the content of the page. Can be optimized, and the search system 100 can improve the efficiency of user search activities.

  The information providing method described in the present embodiment can be realized by executing a program prepared in advance on a computer such as a personal computer or a workstation. This program is recorded on a computer-readable recording medium such as a hard disk, a flexible disk, a CD-ROM, an MO, and a DVD, and is executed by being read from the recording medium by the computer. The program may be a transmission medium that can be distributed via a network such as the Internet.

  The following additional notes are disclosed with respect to the embodiment described above.

(Supplementary Note 1) An extraction process for extracting the publication information posted on the page selected from a set of pages including at least one publication information obtained by giving a search query;
A classification step of classifying the posting information extracted by the extraction step into a plurality of clusters according to the posting content;
Based on the correlation between the page and the cluster classified by the classification step, a calculation step for calculating the degree of suitability of the page for the search query for each page;
Based on the calculation result calculated by the calculation step, a presentation step for presenting page information related to the page group;
An information providing method characterized by including

(Supplementary note 2) including a creation step of creating a directed network graph model with the page and the cluster as nodes,
The calculation step includes
The relevance of the page is calculated based on a transition probability of transition between a node representing the page and a node representing the cluster in the network model created by the creating step. Information providing method described.

(Supplementary note 3)
The information providing method according to claim 2, further comprising: calculating the degree of conformity of the page based on a transition probability of transition between nodes representing the page in the network model.

(Additional remark 4) The information provision method of Additional remark 3 characterized by normalizing the transition probability which changes between the nodes showing the said page.

(Supplementary Note 5) The calculation step includes
From arbitrarily selecting one node from the node group using the transition probability of transitioning between the nodes and the termination probability of ending transition between the nodes, until the transition between the nodes is completed 5. The information providing method according to any one of appendices 2 to 4, wherein the degree of matching of the page is calculated by counting the number of transitions to the node for each node during the period.

(Appendix 6)
The information according to any one of appendices 1 to 5, wherein the posted information is classified into a plurality of clusters based on an appearance frequency of words appearing in the posted information and a similarity between the words. How to provide.

(Additional remark 7) The determination process which determines the word which characterizes the said cluster for every said cluster based on the appearance frequency of the word which appears in the publication information classified into the said cluster,
The presenting step includes
The information providing method according to claim 6, further comprising presenting the determination result determined in the determining step as a topic related to the search query.

(Supplementary note 8)
Determining a word that characterizes the cluster based on a value obtained by dividing the number of occurrences of the word appearing in the publication information by the number of occurrences of the word appearing in all the publication information published in the page group. The information providing method according to appendix 7, which is a feature.

(Supplementary note 9)
9. The information providing method according to appendix 7 or 8, wherein a word characterizing the cluster is determined based on an appearance frequency of a word with a specific part of speech appearing in the posted information.

(Additional remark 10) The topic presentation process which presents the determination result determined by the said determination process as the said topic,
A selection step for accepting selection of an arbitrary topic from among a plurality of topics presented by the topic presentation step,
The calculation step includes
The page is counted by counting the number of transitions made to each node between the time when the node representing the cluster characterized by the topic selected by the selection step is selected and the time when the transition between the nodes ends. The information provision method according to any one of appendices 7 to 9, wherein the degree of fitness of the information is calculated.

(Supplementary Note 11) Any one of Supplementary Notes 1 to 10, wherein the page group includes at least one page that does not have a link relationship with another page included in the page group. Information providing method described in one.

(Supplementary Note 12) Extraction means for extracting the publication information posted on the page selected from the set of pages including at least one publication information obtained by giving a search query;
Classification means for classifying the publication information extracted by the extraction means into a plurality of clusters according to the contents of the publication;
Calculation means for calculating the degree of suitability of the page for the search query for each page based on the correlation between the page and the clusters classified by the classification means;
Presenting means for presenting page information related to the page group based on the calculation result calculated by the calculating means;
An information providing apparatus comprising:

(Supplementary note 13)
An extraction means for extracting the publication information posted on the selected page from a set of pages including at least one publication information obtained by giving a search query;
Classification means for classifying the posting information extracted by the extraction means into a plurality of clusters according to the posting content,
Calculation means for calculating, for each page, the degree of suitability of the page with respect to the search query based on the correlation between the page and the cluster classified by the classification means;
Presenting means for presenting page information related to the page group based on the calculation result calculated by the calculating means;
An information providing program characterized by functioning as

(Supplementary Note 14) A computer-readable recording medium on which the information providing program according to Supplementary Note 12 is recorded.

It is a system configuration figure of a search system. It is explanatory drawing which shows the hardware constitutions of a computer apparatus. It is explanatory drawing which shows the outline | summary of this Embodiment. It is explanatory drawing which shows the memory content of a search result table. It is explanatory drawing which shows the memory content of an article link table. It is explanatory drawing which shows the memory content of a blog link table. It is explanatory drawing which shows the memory content of an analysis result table. It is a block diagram which shows the functional structure of an information provision apparatus. It is explanatory drawing which shows an example of an analysis result. It is explanatory drawing which shows the memory content of an article cluster table. It is explanatory drawing which shows the memory content of a blog cluster table. It is a directed graph which shows an example of a network model. It is explanatory drawing which shows the memory content of a network table. It is explanatory drawing which shows the memory content of a score table. It is explanatory drawing which shows the memory content of a feature word table. It is explanatory drawing (the 1) which shows the example of a screen displayed on a display. It is explanatory drawing (the 2) which shows the example of a screen displayed on a display. It is explanatory drawing which shows an example of the search procedure in a client terminal. It is a flowchart which shows an example of the creation process procedure which produces a blog link table. It is a flowchart which shows an example of the creation process procedure which produces a blog cluster table. It is a flowchart which shows an example of the creation process procedure which produces a network table. It is a flowchart which shows an example of the calculation processing procedure which calculates the fitness of a blog. It is a flowchart which shows an example of the creation process procedure which produces a feature word table. It is a flowchart which shows an example of the specific process sequence of a weighting process. It is a flowchart which shows an example of the presentation process procedure of a search result.

DESCRIPTION OF SYMBOLS 100 Search system 101 Information provision apparatus 102 Database server 103,103-1 to 103-n Client terminal 400 Search result table 500 Article link table 600 Blog link table 700 Analysis result table 801 Reception part 802 Search part 803 Extraction part 804 Classification part 805 Calculation unit 806 Creation unit 807 Presentation unit 808 Determination unit 809 Selection unit 1000 Article cluster table 1100 Blog cluster table 1200 Network model 1300 Network table 1400 Score table 1500 Feature word table 1600, 1700, 1810, 1820 Search results

Claims (3)

Display each set name of multiple sets formed according to each posting content of multiple web pages searched according to the search query,
Displaying title information of web pages belonging to the set of the selected set names among the set names of the plurality of displayed sets,
An information providing program for causing a computer to execute processing. Display each set name of a plurality of sets formed according to each posting content of a plurality of web pages searched according to the search query, and select from among the set names of the displayed plurality of sets A control unit that displays title information of web pages belonging to a set of set names,
An information providing apparatus. A method for providing a search service executed by a server that outputs a search result to the client in response to a search request from a client,
Providing each client with a set name of each of a plurality of sets formed according to the contents of each of a plurality of web pages searched according to a search query transmitted from the client;
When any one of the set names of the plurality of sets is selected by the client, the title information of the web page belonging to the selected set of the set names is provided to the client.
A search service providing method executed by a server.

Images