JP2008242689A - Content classification system, server, terminal device, program, and recording medium - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a content classification system capable of classifying content information at a practical classifying accuracy regardless of the amount of teacher data on the user side, a server, a terminal device, a program, and a recording medium. <P>SOLUTION: A classification rule information storage part 24 of a server 2 stores server-oriented classification rule information that is based on the result of classifying content information by a plurality of users. A communication part 20 sends the server-oriented classification rule information to a user terminal 3. A communication part 30 of the user terminal 3 receives the server-oriented classification rule information from the server 2. A classification rule information storage part 34 stores the received server-oriented classification rule information as user-oriented classification rule information. A news classifying part 35 classifies news data according to the user-oriented classification rule information. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a technique for classifying content information based on classification rule information indicating a classification rule for content information such as net news.

The operation of the content information automatic classification system is composed of two steps: generation of classification rule information and classification of content information based on the classification rule information. A sufficient amount of teacher data (learning data) is required to generate the classification rule information. For example, an automatic classification system represented by PHPNaiveBaysianFilter (see Non-Patent Document 1) generates classification rule information using teacher data by the user himself and performs automatic classification based on the classification rule information.
“PHP Naive Bayesian Filter”, [online], [Search February 23, 2007], Internet <URL: http://www.xhtml.net/php/PHPNaiveBayesianFilter>

  However, the conventional automatic classification system has a problem that practical classification accuracy cannot be realized without a large amount of teacher data by the user.

  The present invention has been made in view of the above-described problems, and is a content classification system, a server, and a server capable of performing classification of content information with practical classification accuracy regardless of the amount of user-side teacher data. It is an object to provide a terminal device, a program, and a recording medium.

  The present invention has been made in order to solve the above-described problem. A server that stores classification rule information indicating a classification rule of the content information based on a result of classification of content information by a plurality of users, A content classification system including a terminal device that executes classification, wherein the server is a server-side classification rule information storage unit that stores the classification rule information, and a server side that transmits the classification rule information to the terminal device A classification rule information transmitting means, wherein the terminal device has received a terminal side content information storage means for storing the content information, a terminal side classification rule information receiving means for receiving the classification rule information from the server, Terminal-side classification rule information storage means for storing the classification rule information, and the information stored in the terminal-side classification rule information storage means. A content classification system characterized in that a terminal-side classifying means for classifying the content information based on the rule information.

  In the content classification system of the present invention, the classification rule information received from the server when the classification rule information is stored in the terminal-side classification rule information storage means before the classification rule information is received from the server. Combining means for combining the classification rule information and the classification rule information stored in the terminal-side classification rule information storage means and storing the new classification rule information in the terminal-side classification rule information storage means as the terminal A device is provided.

  In the content classification system of the present invention, the terminal device updates the classification rule information stored in the terminal-side classification rule information storage unit based on a classification result of the content information; Difference information transmitting means for transmitting difference information indicating a difference of the classification rule information by update to the server, the server receiving difference information receiving means from the terminal device, and the server side classification And a first server-side updating unit that updates the classification rule information stored in the rule information storage unit based on the difference information.

  In the content classification system of the present invention, the terminal device includes category information transmission means for transmitting category information indicating a category desired by a user to the server, and the server receives the category information from the terminal device. Category information receiving means, and search means for searching the classification rule information of a category similar to the category indicated by the category information from the classification rule information stored in the server-side classification rule information storage means. The server-side classification rule information transmitting unit transmits the classification rule information searched by the search unit to the terminal device.

  In the content classification system of the present invention, the terminal device includes terminal-side classification rule information transmitting means for transmitting the classification rule information stored in the terminal-side classification rule information storage means to the server, and the server The server side classification rule information receiving means receives the classification rule information from the terminal device, and the server side classification rule information is received from the classification rule information stored in the server side classification rule information storage means. Search means for searching for the classification rule information similar to the classification rule information received by the means, and the server-side classification rule information transmission means sends the classification rule information searched by the search means to the terminal device. It is characterized by transmitting to.

  In the content classification system of the present invention, the server-side classification rule information storage unit stores the classification rule information for each user cluster.

  In the content classification system according to the present invention, the server may include a server-side content information storage unit that stores the content information, and the content based on the classification rule information stored in the server-side classification rule information storage unit. Server-side classification means for classifying information; and second server-side update means for updating the classification rule information stored in the server-side classification rule information storage means based on the result of classification of the content information. It is characterized by that.

  The present invention also provides a server-side classification rule information storage unit that stores classification rule information indicating a classification rule of the content information based on a result of classification of content information by a plurality of users, and a terminal that executes the classification of the content information A server comprising server-side classification rule information transmitting means for transmitting the classification rule information to an apparatus.

  In the server of the present invention, the terminal device classifies the content information based on the classification rule information, updates the classification rule information based on the classification result of the content information, and updates the classification rule information by the update. When the difference information indicating the difference is transmitted to the server, the difference information receiving means for receiving the difference information from the terminal device and the classification rule information stored in the server-side classification rule information storage means are the difference. And a first server side updating means for updating based on the information.

  In the server of the present invention, the category information receiving means for receiving category information indicating the category desired by the user from the terminal device, and the classification rule information stored in the server-side classification rule information storage means. Search means for searching for the classification rule information of a category similar to the category indicated by the category information, wherein the server-side classification rule information transmission means uses the classification rule information searched by the search means as the terminal device. It is characterized by transmitting to.

  In the server of the present invention, server-side classification rule information receiving means for receiving the classification rule information stored in the terminal-side classification rule information storage means from the terminal device, and the server-side classification rule information storage means Searching means for searching the classification rule information similar to the classification rule information received by the server-side classification rule information receiving means from the stored classification rule information, and the server-side classification rule information The transmission means transmits the classification rule information searched by the search means to the terminal device.

  In the server of the present invention, the server-side classification rule information storage unit stores the classification rule information for each user cluster.

  In the server of the present invention, a server-side content information storage unit that stores the content information, and a server side that classifies the content information based on the classification rule information stored in the server-side classification rule information storage unit It is characterized by comprising: classifying means; and second server side updating means for updating the classification rule information stored in the server side classification rule information storage means based on the classification result of the content information.

  The present invention also provides terminal-side content information storage means for storing content information, and terminal-side classification that receives classification rule information indicating a classification rule for the content information based on the result of classification of content information by a plurality of users from a server. The content information is classified based on the rule information receiving means, the terminal-side classification rule information storage means for storing the received classification rule information, and the classification rule information stored in the terminal-side classification rule information storage means The terminal device is characterized by comprising a terminal-side classification means.

  Further, in the terminal device of the present invention, when the classification rule information is stored in the terminal-side classification rule information storage means before the classification rule information is received from the server, the classification rule information received from the server is received. Combining means for combining the classification rule information and the classification rule information stored in the terminal-side classification rule information storage means and storing the new classification rule information in the terminal-side classification rule information storage means is provided. It is characterized by.

  Further, in the terminal device of the present invention, terminal side update means for updating the classification rule information stored in the terminal side classification rule information storage means based on a classification result of the content information, and the classification rule information by update And difference information transmitting means for transmitting difference information indicating the difference to the server.

  The terminal device of the present invention is characterized by comprising category information transmitting means for transmitting category information indicating a category desired by the user to the server.

  The terminal device according to the present invention further includes terminal-side classification rule information transmitting means for transmitting the classification rule information stored in the terminal-side classification rule information storage means to the server.

  Moreover, this invention is a program for functioning a computer as said server.

  Moreover, this invention is a program for functioning a computer as said terminal device.

  The present invention is a computer-readable recording medium on which the above program is recorded.

  According to the present invention, the classification rule information is transmitted from the server to the terminal device, and the terminal device can use the classification rule information for content information classification. Therefore, regardless of the amount of teacher data in the terminal device. Thus, it is possible to obtain an effect that the content information can be classified with practical classification accuracy.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows the configuration of a news classification system (content classification system) according to an embodiment of the present invention. The news classification system includes a news distribution device 1, a server 2, and a user terminal 3 (terminal device), which are connected to a network 4. The news distribution device 1 distributes news data via the network 4. The server 2 manages the classification rule information indicating the news data classification rule, and updates its own classification rule information based on the news data classification result by the user terminal 3 and the news data classification result by the server 2 itself. The user terminal 3 is a terminal owned by the user. Although only one user terminal 3 is illustrated in FIG. 1, it is assumed that there are a plurality of user terminals similar to the user terminal 3.

  In the server 2, the communication unit 20 communicates with the news distribution apparatus 1 and the user terminal 3 via the network 4. The news storage unit 21 stores the news data received from the news distribution device 1 by the communication unit 20 in the news storage unit 22. The news storage unit 22 stores news data. The classification rule information management unit 23 manages the classification rule information stored in the classification rule information storage unit 24. Hereinafter, the classification rule information managed by the server 2 is referred to as server classification rule information.

  The classification rule information storage unit 24 stores server classification rule information and a synonym dictionary described later. The news classification unit 25 classifies the news data based on the server classification rule information. The news data is classified as a category label (car, gourmet, business, entertainment, etc.). The news storage unit 22 and the classification rule information storage unit 24 may be configured by separate recording media, or may be configured by separate recording areas in the same recording medium.

  In the user terminal 3, the communication unit 30 communicates with the news distribution apparatus 1 and the server 2 via the network 4. The news storage unit 31 stores the news data received from the news distribution device 1 by the communication unit 30 in the news storage unit 32. The news storage unit 32 stores news data. The classification rule information management unit 33 manages the classification rule information stored in the classification rule information storage unit 34. Hereinafter, the classification rule information managed by the user terminal 3 is referred to as user classification rule information.

  The classification rule information storage unit 34 stores user classification rule information. The news classification unit 35 classifies the news data based on the user classification rule information. In the user terminal 3 as well as the server 2, the news data is classified as being associated with the category label. The information input unit 36 is input by the user when the user manually categorizes the news data, as described later, or when inputting a desired category label for acquiring server classification rule information from the server 2. Process information. The news storage unit 32 and the classification rule information storage unit 34 may be configured by separate recording media, or may be configured by separate recording areas in the same recording medium.

  Next, data used in the present embodiment will be described. FIG. 2 shows an example of news data in the news distribution apparatus 1. News data managed by the news distribution apparatus 1 is periodically updated by a news provider (newspaper company or the like). The news ID 200 is an identification code given to each news. Each news ID is not necessarily unique. The news title 210 and the news body 220 are text data.

  FIG. 3 shows an example of user data managed by the server 2. The user ID 300 is an identification code given to a user who uses the news classification system. The cluster ID 310 is the ID of the cluster to which the user belongs. Each user belongs to one of a plurality of user clusters (user groups). The user cluster includes a male cluster and a female cluster. The preference between users belonging to the same user cluster is closer than the preference between users belonging to different user clusters.

  FIG. 4 shows an example of news data in the server 2. News ID 400 and news title 410 are the same as news ID 200 and news title 210 shown in FIG. The word 420 is a word included in the news text. When news data received from the news distribution device 1 is stored in the news storage unit 22, the news storage unit 21 extracts words from the data of the news text.

  FIG. 5 shows an example of a synonym dictionary in the server 2. One representative word 500 is associated with one or more related words 510 having a meaning related to the representative word 500. A specific method of using the synonym dictionary will be described later.

  6 and 7 show an example of server classification rule information in the server 2. The server classification rule information is generated by first combining the user classification rule information of a plurality of users, and then updated based on the news data classification result in the server 2 or the user terminal 3. The server classification rule information in this embodiment is prepared for each user cluster. Further, the server classification rule information of each user cluster is composed of aggregation source data, a word frequency table, and a document frequency table.

  The aggregation source data is data extracted from the news data to be classified, is generated when the news data is classified in the user terminal 3, and is periodically transmitted to the server 2. Although not shown in the figure, the aggregation source data is also generated when the news data is classified in the server 2. FIG. 6A shows the aggregation source data acquired from the user terminal 3 of the user belonging to the user cluster UC1 whose cluster ID is UC1, and FIG. 7A shows the user cluster UC2 whose cluster ID is UC2. The aggregation source data acquired from the user terminal 3 of the user to which the user belongs is shown. The aggregation source data includes user ID 600 (user ID 700), news ID 610 (news ID 710), the frequency of words 620a, 620b, and 620c (words 720a, 720b, and 720c) included in the news body, and the category label 630 of the news data classification destination. (Category label 730).

  The word frequency table shows the frequency of each word included in the news body of news data classified into a predetermined category label. FIG. 6B shows a word frequency table of the user cluster UC1, and FIG. 7B shows a word frequency table of the user cluster UC2. The word frequency table includes a category label 640 (category label 740), words 650a, 650b, 650c, and 650d (words 750a, 750b, and 750c), and a total frequency 660 (total 760) of each word. .

  The document frequency table indicates the number of news data (number of documents) classified into a predetermined category label. FIG. 6C shows a document frequency table of the user cluster UC1, and FIG. 7C shows a document frequency table of the user cluster UC2. The document frequency table includes a category label 670 (category label 770) and a document number 680 (document number 780).

  As shown in FIGS. 6 and 7, the server classification rule information is associated with the category label. Thus, as will be described later, it is possible to present server classification rule information to the user for each category label.

  FIG. 8 shows an example of news data in the user terminal 3. A type 800 indicates a read / unread type of news. A type 810 indicates a news data classification method. When news data classification is performed, “automatic” or “manual” is assigned to the news data as type 810, and when news data classification is not performed, “unclassified” is classified as type 810 in the news data. To be granted. “Automatic” indicates that automatic classification of news data is executed based on the user classification rule information. “Manual” indicates that the news data is finally classified by the manual operation of the user regardless of whether or not the news data is automatically classified based on the user classification rule information.

  The category label 820 indicates a category label to which news data is classified. If the news data has never been classified, “unclassified” is assigned as the category label 820. News ID 830, news title 840, and news body 850 are the same as news ID 200, news title 210, and news body 220 shown in FIG.

  FIG. 9 shows an example of user classification rule information in the user terminal 3. The user classification rule information includes the aggregation source data shown in FIG. 9A, the word frequency table shown in FIG. 9B, the document frequency table shown in FIG. 9C, and the FIG. 9D. It is composed of a difference word frequency table and a difference document frequency table shown in FIG.

  Aggregation source data shown in FIG. 9A is data extracted from news data to be classified, and is generated by the news classification unit 35 when the news data is classified. The aggregation source data includes a news ID 900, the frequencies of words 910a, 910b, and 910c included in the news body, and a category label 920 to which the news data is classified.

  The word frequency table shown in FIG. 9B is the same as the word frequency table shown in FIGS. 6B and 7B. The document frequency table shown in FIG. 9C is the same as the document frequency table shown in FIGS. 6C and 7C. The word frequency table and the document frequency table are updated based on the aggregation source data every time news data is classified.

  The difference word frequency table shown in FIG. 9D and the difference document frequency table shown in FIG. 9E show changes in user classification rule information that occur when the user terminal 3 updates the user classification rule information. The difference information indicating the difference is configured. The difference word frequency table shown in FIG. 9D shows difference information regarding the word frequency table. Further, the difference document frequency table shown in FIG. 9E shows difference information regarding the document frequency table.

  The difference information is transmitted to the server 2 for updating the server classification rule information in the server 2. In this embodiment, difference information is comprised by the difference of the classification rule information for users when the difference information was last transmitted to the server 2, and the latest classification rule information for users. Therefore, the user classification rule information when the difference information was last transmitted to the server 2 is stored in the classification rule information storage unit 34 separately from the latest user classification rule information used for news data classification. The

  Next, a method for managing category labels in the server 2 will be described. In order to suppress the explosive increase in the number of category labels and to associate the category labels uniquely assigned by each user, it is desirable to aggregate the category labels into an appropriate number of category labels. For this reason, in this embodiment, category labels are managed as follows. In the following, two methods will be described as examples of category label management methods.

  First, a method using a synonym dictionary will be described. As described above, a synonym dictionary having a pair of a representative word and a synonym list is stored in the classification rule information storage unit 24 of the server 2. The synonym dictionary is used when a category label given by a user is converted into a category label used for server classification rule information. In that case, the representative word is retrieved using the category label given by the user as a synonym, and the category label given by the user is replaced with the representative word. In order to update the synonym dictionary, for example, words that are not in the synonym dictionary may be separately stored, and the synonym dictionary may be manually updated periodically using this.

  Next, a method for clustering category labels by a statistical method will be described. The category label cluster to which the category label managed by the server 2 belongs and its representative category label are determined by a known statistical method. More specifically, the conditional word frequency distributions of the labels of each pair of category labels are compared with each other, and the distance between the conditional word frequency distributions (Calbachlabler distance or the like) is calculated. Details of the conditional word frequency distribution of the category label will be described when the operation at the time of news data classification by the user terminal 3 is described.

  The distance calculated as described above is used as the similarity between category labels, the category label group is clustered by an arbitrary number of clusters, and a representative category label is determined for each category label cluster. When converting the category label given by the user into the category label used for the server classification rule information, the category label given by the user is converted into the representative category label.

Next, a news data classification method according to the present embodiment will be described. First, the news classification unit 35 of the user terminal 3 (news classification unit 25 of the server 2) calculates the frequency distribution P (c _i ) of the category label c _i (i = 1, 2,..., M). P (c _i ) represents an estimated value of the appearance frequency of each category label document for the entire news data document, and is calculated according to the following equation (1) based on the document frequency table. However, #d ′ is the number of all documents, and #d ′ () is the number of documents satisfying the condition in () (#d ′ (c _i ) is the number of documents classified into the category label c _i ). It is.

Subsequently, the news classification unit 35 (news classification unit 25) _obtains the frequency distribution P (w _j | c _i ) of the conditional word w _j (j = 1, 2,..., N) of the category label c _i. calculate. P (w _j | c _i ) represents an estimated value of the appearance frequency of a specific word in a specific category label, and is calculated according to the following equation (2) based on the word frequency table. However, F is a word vector of the document classified into the category label c _i , | F | is its length, and #d ′ (w _j , c _i ) is classified into the category label c _i , and the word w The number of documents that have _j in the document.

Subsequently, the news classification unit 35 (news classification unit 25) calculates the probability P (c _i , d) that the document d is classified into the category label c _i . P (c _i , d) is calculated according to the following equation (3). Here, N (w _j , d) is the number of words w _{j in} the document d.

Also, the news classification unit 35 (news classification unit 25) calculates a probability P (c _i | d) according to the following equation (4).

After calculating P (c _i | d) for all category labels c _i , the news classification unit 35 (news classification unit 25) determines that P (c _i | d) is the maximum as shown in the following equation (5). This category label c is a category label of the classification destination. Subsequently, the news classification unit 35 (news classification unit 25) reads the news data to be classified stored in the news storage unit 32 (news storage unit 22), associates it with the category label of the classification destination, and stores the news storage unit 32. Store again in (news storage unit 22). Information on the calculation methods of the above-described equations (1) to (5) is also stored in the classification rule information storage unit 34 (classification rule information storage unit 24) as part of the user classification rule information (server classification rule information). ing. A specific classification example by the above-described classification method will be described later.

  Next, the operation of the news classification system according to the present embodiment will be described. First, the operation in which the server 2 searches for server classification rule information suitable for the user will be described with reference to FIG. FIG. 10A shows an operation of searching server classification rule information having a category label similar to the category label desired by the user. FIG. 10B shows an operation for searching server classification rule information similar to the user classification rule information.

  In FIG. 10A, when using the service by the news classification system for the first time, a list of category labels desired by the user (category label list) is input to the information input unit 36 of the user terminal 3 (step S100). . The information input unit 36 outputs the input category label list to the communication unit 30. The communication unit 30 transmits the category label list to the server 2 (step S110).

  The communication unit 20 of the server 2 receives the category label list from the user terminal 3 and outputs it to the classification rule information management unit 23. The classification rule information management unit 23 searches the server classification rule information stored in the classification rule information storage unit 24 for server classification rule information having a category label similar to the category label desired by the user (step S31). S120).

  As an example of this search method, two methods will be described. First, a method of using a category label list and a synonym dictionary will be described. The classification rule information management unit 23 uses the synonym dictionary to replace each element of the category label list desired by the user with the representative word, and the category label list including the representative word becomes the category label of the server classification rule information. The ratio included in the set of is calculated. This ratio is the similarity between the category label list of the user and the category label in the server classification rule information. Then, the classification rule information management unit 23 searches the server classification rule information associated with the category label having a high similarity.

  Next, a method of using a conditional word frequency distribution (see equation (2)) of category labels in the server classification rule information and a word group included therein will be described. The classification rule information management unit 23 collates the word group included in each category label in the server classification rule information with the word group included in the category label list desired by the user, and extracts a common word group. Subsequently, the classification rule information management unit 23 adds the conditional word frequency distribution of the category labels of the common word group, and calculates the similarity between the category label list of the user and the category label in the server classification rule information. To do. Then, the classification rule information management unit 23 searches the server classification rule information associated with the category label having a high similarity.

  Returning to the description of FIG. Subsequent to step S120, the classification rule information management unit 23 displays a category label list made up of category label candidates in order to present to the user the category label candidates of the server classification rule information searched (found by the search). Output to the communication unit 20. The communication unit 20 transmits the category label list to the user terminal 3 (step S130).

  The communication unit 30 of the user terminal 3 receives the category label list from the server 2 and outputs it to the classification rule information management unit 33. The classification rule information management unit 33 executes a process of displaying a category label list on a display unit (not shown) (step S140). The user who confirms the display selects the desired category label and inputs the selection result to the information input unit 36 (step S150). The information input unit 36 outputs information indicating the user selection result to the communication unit 30. The communication unit 30 transmits the information to the server 2 (step S160).

  When the user uses the news classification system for the first time, the user does not yet belong to the user cluster, and the user ID is given to the user after the process shown in FIG. A user cluster is determined.

  Hereinafter, a specific example of the above operation will be described. It is assumed that the user uses the classification by the news classification system for the first time. Assume that “car” and “golf” are input as category labels desired by the user in step S100 of FIG. In step S120, the server 2 searches for category labels in the server classification rules similar to those category labels.

  In the method using the category label list and the synonym dictionary, the search result of the category label is as follows. If the category labels ("car" and "golf") entered by the user are the same as the representative words in the synonym dictionary, there is no need to convert them into representative words. The server 2 calculates the ratio (similarity) included in the set of category labels of the server classification rule information for the category label list desired by the user. FIG. 11A shows the similarity (shown in parentheses) with the category label list desired by the user for each category label of the server classification rule information.

  As a result, the category label list candidates shown in FIG. 11B are displayed on the user terminal 3. The first place of the category label list candidate having a high similarity to the category label desired by the user is the category label list of the user cluster UC1, and the second place of the candidate is the category label list of the user cluster UC2. In each user cluster, the category label list is displayed in descending order of similarity to the category label desired by the user.

  Further, in the method of using the conditional word frequency distribution of the category labels in the server classification rule information and the word groups included therein, the server 2 includes the word groups included in each category label in the server classification rule information and , The conditional frequency distribution of the category labels of the word group common to the word group included in the category label list desired by the user is added, and it is added between the category label list and the category label in the server classification rule information. Similarity.

  FIG. 11C shows the similarity (shown in parentheses) with the category label list desired by the user for each category label of the server classification rule information. Based on the similarity calculation result, category label list candidates are displayed on the user terminal 3. The displayed category label list is the same as that shown in FIG.

  After the category label list candidates are presented as described above, the category label desired by the user is selected from these candidates. For example, the category labels (“car” and “golf”) of the user cluster UC1 shown in FIG. 11 (d) are selected.

  Next, the operation shown in FIG. 10B will be described. The classification rule information management unit 33 of the user terminal 3 reads the user classification rule information from the classification rule information storage unit 34 and outputs it to the communication unit 30. The communication unit 30 transmits the user classification rule information to the server 2 (step S200). The communication unit 20 of the server 2 receives the user classification rule information from the user terminal 3 and outputs it to the classification rule information management unit 23. The classification rule information management unit 23 searches server classification rule information similar to the user classification rule information from the server classification rule information stored in the classification rule information storage unit 24 (step S210).

  At this time, the classification rule information management unit 23 performs conditional word frequency distribution for each category label in the user classification rule information (see equation (2)) and conditional words for each category label in the server classification rule information. Calculate the frequency distribution. Further, the classification rule information management unit 23 compares the calculated conditional word frequency distributions on the user side and the server side, calculates a distance between the conditional word frequency distributions (such as a Cullback library distance), and calculates this distance. The similarity between the user classification rule information and the server classification rule information for each category label. The category labels are presented to the user in the order of the category labels having the highest similarity.

  Subsequently, the classification rule information management unit 23 presents the category label list of the category label candidates to the user in order to present the category label candidates of the searched server classification rule information (found by the search) to the communication unit 20. Output to. The communication unit 20 transmits the classification rule information list to the user terminal 3 (step S220).

  The communication unit 30 of the user terminal 3 receives the category label list from the server 2 and outputs it to the classification rule information management unit 33. The classification rule information management unit 33 executes processing for displaying a category label list on a display unit (not shown) (step S230). The user who confirms the display selects a desired category label, and the selection result is input to the information input unit 36 (step S240). The information input unit 36 outputs information indicating the user selection result to the communication unit 30. The communication unit 30 transmits the information to the server 2 (step S250).

  Hereinafter, a specific example of the above operation will be described. Assume that user classification rule information related to the category labels “car” and “gourmet” is transmitted to the server 2 as the user classification rule information in step S200 of FIG. In step S210, the server 2 searches for server classification rule information similar to the user classification rule information. As a result, the category label list shown in FIG. 12A is presented as a candidate. The user selects, for example, the category labels (“car” and “gourmet”) shown in FIG.

  Next, an operation of acquiring server classification rule information by the user terminal 3 will be described with reference to FIG. The communication unit 20 of the server 2 receives the information transmitted from the user terminal 3 in step S160 or step S250 illustrated in FIG. 10 and outputs the information to the classification rule information management unit 23. The classification rule information management unit 23 extracts server classification rule information corresponding to the user selection result from the server classification rule information stored in the classification rule information storage unit 24 and outputs the server classification rule information to the communication unit 20. (Step S300). The communication unit 20 transmits server classification rule information to the user terminal 3 (step S310).

  The communication unit 30 of the user terminal 3 receives server classification rule information from the server 2 and outputs it to the classification rule information management unit 33. The classification rule information management unit 33 reads out the user classification rule information from the classification rule information storage unit 34, combines the user classification rule information and the server classification rule information, and uses the combined classification rule information for the new user. The classification rule information is stored in the classification rule information storage unit 34. Further, when the user uses the news classification system for the first time and the classification rule information for the user is not stored in the classification rule information storage unit 34, the classification rule information for the server is directly stored as the classification rule information for the user. Stored in the unit 34 (step S320).

  The user classification rule information and the server classification rule information are combined as follows. For the category label, the union of the category labels included in the user classification rule information and the server classification rule information becomes the combined category label. The word frequency in the word frequency table and the document frequency in the document frequency table are as follows. That is, the frequency relating to the category label included in the user classification rule information is used as the frequency after combination relating to the category label included in the user classification rule information and not included in the server classification rule information.

  In addition, the frequency related to the category label included in the server classification rule information is used as the frequency after combination related to the category label included in the server classification rule information and not included in the user classification rule information. In addition, the frequency of combination of category labels included in both the user classification rule information and the server classification rule information (common to both) includes the frequency of the user classification rule information and the frequency of the server classification rule information. What is weighted at a predetermined ratio is used. The weighting ratio can be specified by the user. In the word frequency table, when a category label is included in both the user classification rule information and the server classification rule information, the word belonging to the category label exists only in either the user side or the server side word frequency table The frequency of the word is used as the frequency of the word after synthesis. However, as described below, for the category label common to the user classification rule information and the server classification rule information, the server classification rule information of the category label may not be the target of synthesis.

  For the category labels common to the user classification rule information and the server classification rule information, the word frequency distribution on the user side and the server side (the total frequency of each category label is summed and normalized) (The one divided by the number) is similar), the server classification rule information of the category label becomes the synthesis target. When the word frequency distribution is not similar between the user side and the server side, the server classification rule information is excluded from the synthesis targets for the category label.

  Hereinafter, a specific example of the above operation will be described. It is assumed that the user U1 having the user ID “U1” selects only the server and the gourmet-related server classification rule information from the server classification rule information of the user cluster UC1. 14A and 14B show a word frequency table and a document frequency table acquired by the user terminal 3 from the server 2, respectively. The word frequency table shown in FIG. 14A is a word frequency table of category labels (“car” and “gourmet”) selected by the user from the server word frequency table shown in FIG. Extracted. Also, the document frequency table shown in FIG. 14B is the document frequency of the category label (“car” and “gourmet”) selected by the user from the server document frequency table shown in FIG. 6C. The table is extracted.

  The word frequency table and the document frequency table stored in the classification rule information storage unit 34 of the user terminal 3 before obtaining the word frequency table and the document frequency table from the server 2 are shown in FIGS. 9B and 9C, respectively. It shall be as it was. FIG. 15A shows a word frequency table after synthesis. For each word of each category label, a value obtained by adding the frequency in the word frequency table shown in FIG. 14A and the frequency in the word frequency table shown in FIG. The frequency is within. FIG. 15B shows a document frequency table after synthesis. For each category label, a value obtained by adding the frequency (number of documents) in the word frequency table shown in FIG. 14B and the frequency (number of documents) in the document frequency table shown in FIG. This is the frequency (number of documents) in the later document frequency table.

  Next, the news data classification operation by the user terminal 3 will be described with reference to FIG. When news data is transmitted from the news distribution apparatus 1 (step S400), the communication unit 30 of the user terminal 3 receives the news data and outputs it to the news storage unit 31. The news storage unit 31 stores the news data in the news storage unit 32 (step S410).

  Subsequently, the news classification unit 35 reads news data from the news storage unit 32. The news classification unit 35 extracts words from the news body data included in the read news data, calculates the frequency of the words, and generates aggregation source data. Further, the news classification unit 35 reads the user classification rule information from the classification rule information storage unit 34, and executes news data classification based on the aggregation source data and the user classification rule information (step S420). The aggregation source data is stored in the classification rule information storage unit 34 after adding the category label information of the classification destination after the classification label category label of the news data is determined.

  Hereinafter, specific examples of classification of news data will be described. In the following description, the word frequency table and the document frequency table shown in FIG. 15 are used as the user classification rule information. FIG. 17A shows news data to be classified. FIG. 17 (b) shows the aggregation source data generated from the news data.

The probability P (car | N5) (the above P (c _i | d)) that the news data “N5” is classified into the category label “car” is calculated as follows. The frequency distribution P (vehicle) (the above P (c _i )) of the category label “car” is expressed by the following equation (6).
P (car) = (103 + 1) / (607 + 2) = 0.17 (6)

Also, the frequency distributions P (car | car), P (benz | car), P (french | car), P for the words "car", "benz", "French", and "Japanese food" included in the news text. (Japanese food | car) (above mentioned P (w _j | c _i )) is expressed by the following equations (7) to (10).
P (automobile | car) = (502 + 1) / (527 + 4) = 0.95 (7)
P (Benz | car) = (23 + 1) / (527 + 4) = 0.05 (8)
P (French | Car) = (2 + 1) / (527 + 4) = 0.01 (9)
P (Japanese food | car) = (1 + 1) / (527 + 4) = 0.00 (10)

Therefore, the probability P (car, N5) is expressed by the following equation (11).
P (car, N5) = P (car) x P (car | car) ⁰ xP (Benz | car) ⁰ xP (French | car) ⁰ xP (Japanese food | car) ²
≒ 0.0% (11)

Similarly, the probability P (dish, N5) is expressed by the following equation (12).
P (Cooking, N5) = P (Cooking) x P (Automobile | Cooking) ⁰ x P (Benz | Cooking) ⁰ x P (French | Cooking) ⁰ x P (Japanese cooking | Cooking) ²
≒ 20.5% (12)

  Further, in this case, the above-described equation (5) becomes the following equation (13). From this equation (13) and P (car, N5) <P (dish, N5), the news data “N5” is classified into the category label “dish”.

  Hereinafter, an example of a screen displayed on a display unit (not shown) in the user terminal 3 will be described. FIG. 18A shows a screen for automatically classifying unread news data. When the news data is received, the news data is temporarily stored in the classification folder of the news storage unit 32 and then stored in the unread folder of the news storage unit 32 by automatic classification. A category label 1800 and news data 1810 classified into the category label 1800 are displayed on the screen as the contents of the unread folder.

  Of the news data stored in the unread folder, the data that the user desires to store is stored in the storage folder of the news storage unit 32. A category label 1820 and news data 1830 classified into the category label 1820 are displayed on the screen as the contents of the storage folder. The user can create a new category label in the storage folder. The category label of the unread folder is synchronized with the storage folder.

  When the user clicks news data 1810 with a mouse or the like, the content of the news is displayed on news display screen 1840. When discarding the displayed news data, the user right-clicks the display screen 1840 with a mouse or the like and selects “Discard” from the pull-down menu that is subsequently displayed. Deleted. When the news data is stored with the automatically classified category label, the user right-clicks the display screen 1840 with a mouse or the like, and “Save to XXX” (“ If “XXX” indicates the category label of the automatic classification destination, news data is stored in the area of the category label “XXX” in the storage folder of the news storage unit 32.

  Manual classification can also be performed by changing automatically classified category labels. When storing news data by manual classification, the user clicks the display screen 1840 with a mouse or the like and drags it to a desired category label in the storage folder, so that the news data is stored in the storage folder of the news storage unit 32. Stored in the category label area.

  FIG. 18B shows a screen for automatically classifying already read news data. When the news data is received, the news data is temporarily stored in an unread folder of the news storage unit 32, and then stored in a storage folder of the news storage unit 32 by automatic classification or manual classification after browsing. On the screen, uncategorized news data 1850 is displayed as the contents of the unread folder.

  Of the news data stored in the unread folder, the data that the user desires to store is stored in the storage folder of the news storage unit 32. A category label 1860 and news data 1870 classified into the category label are displayed on the screen as the contents of the storage folder. The user can create a new category label in the storage folder.

  When the user clicks news data 1850 with a mouse or the like, the content of the news is displayed on news display screen 1880. When discarding the displayed news data, the user right-clicks the display screen 1880 with a mouse or the like and selects “Discard” from the pull-down menu that is subsequently displayed. Deleted. When news data is stored by automatic classification, the user right-clicks the display screen 1880 with a mouse or the like, and “Save to XXX” (“XXX” is an automatic classification destination) from the pull-down menu displayed subsequently. News data is stored in the area of category label “XXX” in the storage folder of the news storage unit 32.

  Manual classification is also possible. When news data is stored by manual classification, the user clicks the display screen 1880 with a mouse or the like and drags it to a desired category label in the storage folder, so that the news data is stored in the storage folder of the news storage unit 32. Stored in the category label area.

  Next, the update operation of the user classification rule information by the user terminal 3 will be described with reference to FIG. After the news data is classified by the automatic classification or the manual classification, the classification rule information management unit 33 displays the aggregation source data indicating the frequency of words included in the classified news data and the user classification rule information as a classification rule information storage unit. 34, the user classification rule information is updated based on the aggregation source data, and the updated user classification rule information is stored in the classification rule information storage unit 34 (step S500). More specifically, the classification rule information management unit 33 updates the frequency of each word included in the word frequency table based on the frequency of each word included in the aggregation source data, and the frequency (document Number).

  Subsequently, the classification rule information management unit 33 displays the classification rule information for the user and the latest classification rule information for the user when the change (difference information) of the classification rule information for the user is finally notified to the server 2. It reads from the memory | storage part 34, produces | generates the difference information which shows those differences, and outputs it to the communication part 30 (step S510). This difference information includes the above-described difference word frequency table, difference document frequency table, and user ID. The communication unit 30 transmits the difference information to the server 2 (Step S520).

  Hereinafter, a specific example of the above operation will be described. It is assumed that the user terminal 3 of the user U1 having the user ID “U1” has classified news data having the news ID “N5” by manual classification. Further, it is assumed that the word frequency table and the document frequency table before the update are as shown in FIGS. 15 (a) and 15 (b), respectively. Further, it is assumed that the word frequency table and the document frequency table when the change of the user classification rule information is finally notified to the server 2 are as shown in FIGS. 15 (a) and 15 (b), respectively. FIG. 20A shows the news data after manual classification. A type 2000 indicating unread / read indicates “read”, and a type 2010 indicating a classification method indicates “manual”. The category label 2020 indicates that the news data is manually classified as “dishes”.

  FIG. 20B shows the aggregation source data stored in the classification rule information storage unit 34. The word frequency table and the document frequency table are updated based on the total data. FIG. 20C shows an updated word frequency table, and FIG. 20D shows an updated document frequency table. In the aggregation source data shown in FIG. 20B, the frequency of the word “Japanese food” is 2, and the frequency of other words is 0. The news data is classified into the category label “Cooking”. Therefore, as can be seen by comparing the word frequency tables shown in FIG. 15A and FIG. 20C, 2 is added to the frequency of the word “Japanese food” of the category label “Cooking” (also in the total). 2 is added). Further, as can be seen by comparing the document frequency tables shown in FIG. 15B and FIG. 20D, 1 is added to the number of documents of the category label “Cooking” (1 is also added to the total). )

  The difference between the frequencies of the word frequency tables shown in FIGS. 15A and 20C is the difference word frequency table. FIG. 21A shows a difference word frequency table. Further, the difference between the frequencies of the document frequency tables shown in FIGS. 15B and 20D is the difference document frequency table. FIG. 21B shows a difference document frequency table.

  Next, the update operation of the server classification rule information by the server 2 will be described with reference to FIG. FIG. 22A shows an operation of updating the server classification rule information based on the difference information received from the user terminal 3. FIG. 22B shows an operation of updating the server classification rule information based on the news data received from the news distribution apparatus 1.

  22A, the communication unit 20 of the server 2 receives the difference information transmitted from the user terminal 3 in step S520 illustrated in FIG. 19 and outputs the difference information to the classification rule information management unit 23. The classification rule information management unit 23 determines the user cluster to which the user belongs based on the user ID included in the difference information (step S600). In this determination, user data (see FIG. 3) in which the user ID and the user cluster ID are associated is also referred to.

  When the user ID that matches the user ID included in the difference information is in the user data, the user cluster having the user cluster ID associated with the user ID is the user cluster to which the user belongs. When the user ID that matches the user ID included in the difference information is not included in the user data, the user cluster to which the user belongs is determined according to the similarity between the category labels described above.

  Following the determination of the user cluster, the classification rule information management unit 23 reads the server classification rule information of the user cluster determined to belong to the user from the classification rule information storage unit 24, and based on the difference information from the user terminal 3. The server classification rule information is updated, and the updated server classification rule information is stored in the classification rule information storage unit 24 (step S610). More specifically, the classification rule information management unit 23 updates the frequency of each word included in the word frequency table based on the frequency of each word included in the difference word frequency table, and is included in the difference document frequency table. The frequency (number of documents) in the document frequency table is updated based on the frequency (number of documents). This is the same as the operation of updating the user classification rule information on the user terminal 3 based on the server classification rule information received from the server 2. However, as described below, even for the category label common to the difference information and the server classification rule information, the difference information may or may not be directly combined.

  For category labels common to difference information and server classification rule information, the word frequency distribution (the frequency of each word in each category label is summed and normalized (divided by the total number of words in each category label). ))) Are similar, the difference information becomes the synthesis target as it is. On the other hand, when the word frequency distribution is not similar between the user side and the server side, a new category label is added to the server classification rule information, and the information included in the difference information is left as it is for the server for the category label. This is classification rule information.

  Hereinafter, a specific example of the above operation will be described. It is assumed that the user terminal 3 of the user U1 having the user ID “U1” has transmitted the difference information to the server 2. Assume that the difference word frequency table and the difference document frequency table transmitted from the user terminal 3 are as shown in FIGS. 21 (a) and 21 (b), respectively. Further, it is assumed that the word frequency table and the document frequency table of the server classification rule information before the update are as shown in FIGS. 6B and 6C, respectively. In the determination of the user cluster in step S600, it is determined from the contents of the user data shown in FIG. 3 that the user U1 belongs to the user cluster UC1. Therefore, the user classification rule information of the user cluster UC1 is updated.

  In the word frequency table shown in FIG. 21A, the frequency of the word “Japanese food” of the category label “Cooking” is 2, and the frequency of the other words is 0. For this reason, only the frequency of the word of the category label “cooking” is reflected in the update. The user-side category label “cooking” is converted into the representative word “gourmet” in the server 2 using a synonym dictionary. When the word frequency distribution is almost the same on the user side and the server side for the category label “gourmet”, the updated word frequency table is as shown in FIG. 23A, and the updated document frequency table is shown in FIG. )become that way. As can be seen by comparing the word frequency tables shown in FIG. 6B and FIG. 23A, 2 is added to the frequency of the word “Japanese food” of the category label “Gourmet” (2 is also added to the total). Have been). Further, as can be seen by comparing the document frequency tables shown in FIG. 6C and FIG. 23B, 1 is added to the number of documents of the category label “Gourmet” (1 is also added to the total). )

  For the category label “gourmet”, when the word frequency distribution is different between the user side and the server side, the updated word frequency table is as shown in FIG. 23C, and the updated document frequency table is shown in FIG. )become that way. As can be seen by comparing the word frequency tables shown in FIG. 6B and FIG. 23C, the category label “Gourmet (2)” is added, and the frequency of the word “Japanese food” is 2. Further, as can be seen by comparing the document frequency tables shown in FIG. 6C and FIG. 23D, the category label “Gourmet (2)” is added and the number of documents is 1 (the total is further). Is added to 1).

  Next, the operation shown in FIG. 22B will be described. When news data is transmitted from the news distribution apparatus 1 (step S700), the communication unit 20 of the server 2 receives the news data and outputs it to the news storage unit 21. The news storage unit 21 stores the news data in the news storage unit 22 (step S710).

  Subsequently, as in step S420 of FIG. 16, the news classification unit 25 executes news data classification (step S720). Subsequently, as in step S500 of FIG. 19, the classification rule information management unit 23 updates the server classification rule information (step S730).

  Hereinafter, a specific example of the above operation will be described. It is assumed that the word frequency table and the document frequency table before update are as shown in FIGS. 23 (a) and 23 (b), respectively. FIG. 24A shows news data to be classified. FIG. 24 (b) shows the aggregation source data generated from the news data. The word frequency table and the document frequency table are updated based on the total data.

  FIG. 24C shows the updated word frequency table, and FIG. 24D shows the updated document frequency table. In the aggregation source data shown in FIG. 24B, the frequency of the word “French” is 3, and the frequency of the other words is 0. The news data is classified into the category label “Gourmet”. Therefore, as can be seen by comparing the word frequency tables shown in FIGS. 23A and 24C, 3 is added to the frequency of the word “French” of the category label “Gourmet” (also in the total) 3 is added). Further, as can be seen by comparing the document frequency tables shown in FIGS. 23B and 24D, 1 is added to the number of documents of the category label “Gourmet” (1 is also added to the total). )

  Next, a user cluster generation operation and an update operation by the server 2 will be described with reference to FIG. As an initial state, it is assumed that the classification rule information for N users (N is a plurality) is stored in the classification rule information storage unit 24. The classification rule information management unit 23 generates a user cluster based on the user classification rule information (step S800).

  The user cluster is generated as follows. First, the classification rule information management unit 23 sums and normalizes the frequency of each word in all category labels in the user classification rule information for each user classification rule information (total words in the user classification rule information). Generate user word frequency distribution (divided by number). Subsequently, the classification rule information management unit 23 compares the word frequency distributions of different users and calculates the distance between the word frequency distributions (such as a Cullback library distance). The classification rule information management unit 23 uses this distance as a similarity between users, and clusters a user group with an arbitrary number of clusters. As a result, users with high similarity belong to the same user cluster.

  Further, the classification rule information management unit 23 adds and normalizes the word groups in the user classification rules of each user included in the user cluster (by dividing by the total number of words included in the user cluster) to thereby increase the word frequency of the user cluster. Generate a distribution. Further, the classification rule information management unit 23 synthesizes user classification rule information for users belonging to each user cluster, and generates initial server classification rule information for each user cluster.

  Subsequently, the classification rule information management unit 23 updates the user cluster at an arbitrary timing (step S810). The user cluster update operation is the same as the user cluster generation operation. However, the server 2 generates the word frequency distribution based on the aggregation source data received from the user terminal 3 and the aggregation source data generated when the server 2 classifies the news data. As the user cluster is updated, the server classification rule information of each user cluster is also updated, but the update is also performed based on the aggregation source data.

  As described above, according to the present embodiment, server classification rule information is transmitted from the server 2 to the user terminal 3, and the user terminal 3 can use the server classification rule information for news data classification. Therefore, news data can be classified with practical classification accuracy regardless of the amount of teacher data in the user terminal 3. In particular, even when the user terminal 3 does not have user classification rule information, the news data can be classified using the server classification rule information.

  Further, when the user terminal 3 has user classification rule information in advance, the server classification rule information of the server 2 and its own user classification rule information are combined to form new user classification rule information. Thereby, the classification accuracy of news data in the user terminal 3 can be improved.

  In addition, since the server 2 updates its server classification rule information based on the difference information transmitted from the user terminal 3, the classification result by the user terminal 3 is reflected in the server classification rule information. The server classification rule information can be updated while following the user's preference.

  Further, the server 2 searches the server classification rule information of the category label similar to the category label desired by the user, and transmits the server classification rule information found by the search to the user terminal 3, whereby the user terminal 3 Since the server classification rule information closer to the user's preference can be acquired, the accuracy of news data classification in the user terminal 3 can be improved.

  Similarly, the server 2 searches the server classification rule information similar to the user classification rule information, and transmits the server classification rule information found by the search to the user terminal 3, whereby the user terminal 3 Since it becomes possible to acquire server classification rule information closer to the preference, the accuracy of news data classification in the user terminal 3 can be improved.

  Further, the server 2 possesses server classification rule information for each user cluster, whereby server classification rule information suitable for each user group having different preferences can be provided.

  In addition, the server 2 can update server classification rule information based on its own classification result, thereby providing server classification rule information suitable for the classification of news data updated as needed.

  As described above, the embodiments of the present invention have been described in detail with reference to the drawings. However, the specific configuration is not limited to the above-described embodiments, and includes design changes and the like without departing from the gist of the present invention. .

It is a block diagram which shows the structure of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure which shows the news data in a news delivery apparatus. FIG. 6 is a reference diagram showing user data managed by a server in an embodiment of the present invention. In one Embodiment of this invention, it is a reference figure which shows the news data in a server. In one Embodiment of this invention, it is a reference figure which shows the synonym dictionary in a server. It is a reference figure which shows the classification rule information for servers in one Embodiment of this invention. It is a reference figure which shows the classification rule information for servers in one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure which shows the news data in a user terminal. It is a reference figure which shows the classification rule information for users in one Embodiment of this invention. It is a sequence diagram which shows the procedure of operation | movement (search operation | movement of the classification rule information for servers by the server 2) of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure which shows the category label list | wrist of the classification rule information for servers searched by the server. In one Embodiment of this invention, it is a reference figure which shows the category label list | wrist of the classification rule information for servers searched by the server. It is a sequence diagram which shows the procedure of the operation | movement (acquisition operation | movement of the classification rule information for servers by a user terminal) of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure which shows the classification | category rule information for servers which the user terminal acquired from the server. FIG. 6 is a reference diagram showing user classification rule information generated (updated) by combining user classification rule information and server classification rule information in an embodiment of the present invention. It is a sequence diagram which shows the procedure of the operation | movement (news data classification operation by a user terminal) of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure for demonstrating the classification | category operation | movement of the news data by a user terminal. In one Embodiment of this invention, it is a reference figure which shows the display screen of the user terminal 3. FIG. It is a sequence diagram which shows the procedure of the operation | movement (update operation | movement of the classification rule information for users by a user terminal) of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure for demonstrating the update operation | movement of the classification rule information for users by a user terminal. In one Embodiment of this invention, it is a reference figure which shows the difference information by a user terminal. It is a sequence diagram which shows the procedure of the operation | movement (update operation | movement of the server classification rule information by a server) of the news classification system by one Embodiment of this invention. In one Embodiment of this invention, it is a reference figure for demonstrating the update operation | movement of the classification rule information for servers by the server. In one Embodiment of this invention, it is a reference figure for demonstrating the update operation | movement of the classification rule information for servers by the server. It is a sequence diagram which shows the procedure of the operation | movement (The production | generation operation | movement and update operation of a user cluster by a server) of the news classification system by one Embodiment of this invention.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... News distribution apparatus, 2 ... Server, 3 ... User terminal, 4 ... Network, 20, 30 ... Communication part, 21, 31 ... News preservation | save part, 22, 32. ..News storage unit, 23, 33 ... Classification rule information management unit, 24, 34 ... Classification rule information storage unit, 25, 35 ... News classification unit, 36 ... Information input unit

Claims (21)

A content classification system comprising: a server that stores classification rule information indicating a classification rule of the content information based on a result of classification of content information by a plurality of users; and a terminal device that executes the classification of the content information,
The server
Server-side classification rule information storage means for storing the classification rule information;
Server-side classification rule information transmitting means for transmitting the classification rule information to the terminal device,
The terminal device
Terminal-side content information storage means for storing the content information;
Terminal-side classification rule information receiving means for receiving the classification rule information from the server;
Terminal-side classification rule information storage means for storing the received classification rule information;
A content classification system comprising: terminal side classification means for classifying the content information based on the classification rule information stored in the terminal side classification rule information storage means.   When the classification rule information is stored in the terminal-side classification rule information storage means before the classification rule information is received from the server, the classification rule information received from the server and the terminal-side classification rule The terminal device includes combining means for combining the classification rule information stored in the information storage means and storing the new classification rule information in the terminal-side classification rule information storage means. Item 4. The content classification system according to Item 1. The terminal device
Terminal-side update means for updating the classification rule information stored in the terminal-side classification rule information storage means based on the classification result of the content information;
Difference information transmission means for transmitting difference information indicating the difference of the classification rule information by update to the server,
The server
Difference information receiving means for receiving the difference information from the terminal device;
The server-side classification rule information storage means comprises first server-side update means for updating the classification rule information stored based on the difference information. The content classification system described. The terminal device includes category information transmission means for transmitting category information indicating a category desired by a user to the server,
The server
Category information receiving means for receiving the category information from the terminal device;
Search means for searching the classification rule information of a category similar to the category indicated by the category information from the classification rule information stored in the server-side classification rule information storage means,
The content classification system according to any one of claims 1 to 3, wherein the server-side classification rule information transmission unit transmits the classification rule information searched by the search unit to the terminal device. The terminal device comprises terminal-side classification rule information transmitting means for transmitting the classification rule information stored in the terminal-side classification rule information storage means to the server,
The server
Server-side classification rule information receiving means for receiving the classification rule information from the terminal device;
Search means for retrieving the classification rule information similar to the classification rule information received by the server-side classification rule information receiving means from among the classification rule information stored in the server-side classification rule information storage means; With
The content classification system according to any one of claims 1 to 3, wherein the server-side classification rule information transmission unit transmits the classification rule information searched by the search unit to the terminal device.   6. The content classification system according to claim 1, wherein the server-side classification rule information storage unit stores the classification rule information for each user cluster. The server
Server-side content information storage means for storing the content information;
Server-side classification means for classifying the content information based on the classification rule information stored in the server-side classification rule information storage means;
Second server-side update means for updating the classification rule information stored in the server-side classification rule information storage means based on the classification result of the content information;
The content classification system according to any one of claims 1 to 6, further comprising: Server-side classification rule information storage means for storing classification rule information indicating a classification rule of the content information based on a result of classification of content information by a plurality of users;
Server-side classification rule information transmitting means for transmitting the classification rule information to a terminal device for performing classification of the content information;
A server characterized by comprising: The terminal device classifies the content information based on the classification rule information, updates the classification rule information based on a classification result of the content information, and stores difference information indicating a difference of the classification rule information due to the update. Difference information receiving means for receiving the difference information from the terminal device,
First server side updating means for updating the classification rule information stored in the server side classification rule information storage means based on the difference information;
The server according to claim 8, comprising: Category information receiving means for receiving category information indicating a category desired by the user from the terminal device;
Search means for searching the classification rule information of a category similar to the category indicated by the category information from the classification rule information stored in the server-side classification rule information storage means,
The server according to claim 8 or 9, wherein the server-side classification rule information transmission unit transmits the classification rule information searched by the search unit to the terminal device. Server-side classification rule information receiving means for receiving the classification rule information stored in the terminal-side classification rule information storage means from the terminal device;
Search means for retrieving the classification rule information similar to the classification rule information received by the server-side classification rule information receiving means from among the classification rule information stored in the server-side classification rule information storage means; With
The server according to claim 8 or 9, wherein the server-side classification rule information transmission unit transmits the classification rule information searched by the search unit to the terminal device.   12. The server according to claim 8, wherein the server-side classification rule information storage unit stores the classification rule information for each user cluster. Server-side content information storage means for storing the content information;
Server-side classification means for classifying the content information based on the classification rule information stored in the server-side classification rule information storage means;
Second server-side update means for updating the classification rule information stored in the server-side classification rule information storage means based on the classification result of the content information;
The server according to any one of claims 8 to 12, further comprising: Terminal-side content information storage means for storing content information;
Terminal-side classification rule information receiving means for receiving, from a server, classification rule information indicating a classification rule of the content information based on a result of classification of content information by a plurality of users;
Terminal-side classification rule information storage means for storing the received classification rule information;
Terminal-side classification means for classifying the content information based on the classification rule information stored in the terminal-side classification rule information storage means;
A terminal device comprising:   When the classification rule information is stored in the terminal-side classification rule information storage means before the classification rule information is received from the server, the classification rule information received from the server and the terminal-side classification rule 15. The apparatus according to claim 14, further comprising a combining unit that combines the classification rule information stored in the information storage unit and stores it in the terminal-side classification rule information storage unit as new classification rule information. Terminal equipment. Terminal-side update means for updating the classification rule information stored in the terminal-side classification rule information storage means based on the classification result of the content information;
Difference information transmitting means for transmitting difference information indicating a difference of the classification rule information by update to the server;
16. The terminal device according to claim 14 or 15, further comprising:   The terminal device according to any one of claims 14 to 16, further comprising category information transmission means for transmitting category information indicating a category desired by a user to the server.   The terminal-side classification rule information transmission means for transmitting the classification rule information stored in the terminal-side classification rule information storage means to the server. Terminal equipment.   A program for causing a computer to function as the server according to claim 8.   A program for causing a computer to function as the terminal device according to claim 14.   A computer-readable recording medium on which the program according to claim 19 or 20 is recorded.

Images