JP2014215625A - Data delivering method, data delivering apparatus, computer program, and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enable the user of a terminal, when given a certain set of data, to quickly and readily find a set similar (or related) to the set out of sets of data the user has so far come across.SOLUTION: When a set of data (for instance news data) is delivered to a user terminal, the data is displayed on a display screen 500 of the terminal. Identifying information for identifying sets of data this user has so far come across is stored separately. In accordance with a prescribed instruction from the user terminal regarding the delivered data (for instance an instruction to press the Recall button), sets of data similar to the sets of data covered by the prescribed instruction out of the sets of data the user has so far come across are searched. These searched sets of data are transmitted to and displayed on the user terminal.

Description

  The present invention relates to a data distribution method, a data distribution apparatus, a computer program, and a recording medium for distributing data to a user terminal.

  Currently, the Internet is widely used as a medium for obtaining information. As usage modes, users voluntarily access various homepages, blogs, social networking services (SNS), news sites, etc. from terminals, and articles such as news are regularly distributed to registered users. You can also use the information distribution service.

  Patent Document 1 proposes the following method as an advertisement providing method. That is, a document designated by the user is described by a vector with a plurality of attribute values indicating the characteristics of the document, a vector obtained by combining the vectors of the designated document group is calculated as a vector representing the user's preference, An advertisement that is a candidate for presentation is described as a vector with a plurality of attribute values indicating the characteristics of the advertisement, the similarity between the vector indicating the user's preference and the vector indicating the characteristic of the advertisement is calculated, and the advertisement with a high similarity is prioritized. Present.

  Patent Document 2 proposes an information processing apparatus and method for presenting related information without losing timing according to the situation. In this prior art, the similarity between both is calculated from the inner product of the feature vector of a document corresponding to the occurrence of an event due to mail transmission and reception and the feature vector of each topic (document group). Further, when the total number of words (feature words) of all the topics is n as topic feature vectors, it indicates that the feature vectors of all the topics are represented by vectors in an n-dimensional space. That is, the use of an n-dimensional vector composed of a plurality of word weights is disclosed.

  Patent Document 3 proposes a document processing apparatus that reflects a user's preference in a comparison between documents when calculating a similarity between documents from a document vector representing the characteristics of the document.

  In Patent Literature 4, access information such as a genre and a keyword of a content that the user wants to search, a user ID and a user information obtained by quantifying the user's content preference, and a sum of vectors, There has been proposed a content search apparatus that obtains an inner product of content information vectors obtained by quantifying content attributes and orders the contents based on the calculation result.

JP 2004-118716 A JP 2003-178075 A JP 2000-194723 A JP 2009-245382 A

  By the way, there is a wide variety of information obtained from the Internet, and in particular, a user who uses an information distribution service comes into contact with a large amount of information that is regularly distributed. In such a situation, when a user touches an article, there is a memory of reading an article similar to (or related to) the content of the article in the past. However, it may take time and effort to find the target article, or the search may not be successful.

  In such a background, when the terminal user is given certain data, the present invention quickly and easily finds data similar to (or related to) that data from past data that the user has contacted. A data distribution method, a data distribution apparatus, a computer program, and a recording medium are provided.

  A data distribution method according to the present invention is a data distribution method for distributing data to a user terminal, the step of distributing data to the user terminal, and identifying past data that the user of the user terminal has contacted in the past Data that is similar to the data that has received the predetermined instruction among the past data that the user has contacted in response to a predetermined instruction from the user terminal with respect to the distributed data, And a step of transmitting the detected data to the user terminal.

  In this data distribution method, specific information for specifying past data that the user has contacted in the past is stored, and the data that the user has contacted is used as a trigger (trigger) to resemble the user's past data. Detect data. This makes it possible to automatically estimate and extract data that the user will recall from past data, triggered by the distributed data.

  For this purpose, for example, for data distributed to the user terminal, the method further includes a step of generating a data feature vector representing a feature of each data, and the detecting step includes a data feature vector of the data and a data feature of the past data. By comparing the vectors with each other, the similarity between the data to be compared is obtained, and one or more data having a high similarity are specified from the past data based on the obtained similarity.

  In the step of obtaining the similarity, the data feature vector of the distributed data is compared with the data feature vector of the past data, and a correlation coefficient between both feature vectors can be obtained as the similarity. In addition to the correlation coefficient, the inner product of both feature vectors and the Euclidean distance may be used. Further, the correction may be performed by multiplying the information amount (information entropy) of the data feature vector.

  The method may further include the step of converting the data feature vector into an associative vector obtained by extending the feature, and in the detecting step, the associative vector may be compared with a data feature vector of the past data. As a result, it is possible to supplement the feature of the data that triggers the recall and reduce the leakage of the recall.

  The detecting step may be executed in real time according to the predetermined instruction, or may be executed in advance for each data to be distributed when the data is distributed.

  In the distributing step, the past data obtained by the detecting step executed in advance can be embedded in the distributed data in an initially non-display state and distributed.

  The “article contacted by the user in the past” is at least one of an article read by the user in the past, an article that has been operated by a user representing an intention indicating that the user is interested, and an article that has been received in the past. It is.

  The method may further comprise managing data stored by the user using an external storage service as data that the user has contacted in the past. Thereby, it is possible to expand the range of data to be subjected to the data distribution service.

  A data distribution apparatus according to the present invention is a data distribution apparatus that distributes data to a user terminal, wherein distribution data generation means for generating data to be distributed to the user terminal and a user of the user terminal have contacted the past. Past data storage means for storing specific information for specifying the past data, and the predetermined instruction of the past data contacted by the user in response to a predetermined instruction from the user terminal for the distributed data Data detection means for detecting data similar to the received data, and transmission means for transmitting the detected data to the user terminal.

  A computer program according to the present invention is a computer program for causing a computer to execute a data distribution method for distributing data to a user terminal, the step of distributing data to the user terminal, and a user of the user terminal in the past. In response to a predetermined instruction from the user terminal with respect to the distributed data, a step of storing specific information for specifying the past data that has been contacted, and receiving the predetermined instruction among the past data that the user has contacted A computer program that causes a computer to execute a step of detecting data similar to the detected data and a step of transmitting the detected data to the user terminal.

  The recording medium by this invention is a recording medium which recorded the said computer program so that computer reading was possible.

  According to the present invention, when a terminal user is given certain data, it can quickly and easily resemble (or relate to) that data from the past data that the user has contacted without imposing a burden on the user. ) Can be found.

It is a figure which shows the schematic structure of the whole data delivery system with which the service using the internet in embodiment of this invention is provided. FIG. 2 is a block diagram showing various functions of the terminal shown in FIG. 1. It is a block diagram showing the various functions of the service server shown in FIG. It is a block diagram for demonstrating the function of the whole data delivery system in embodiment of this invention. It is a figure which shows the conceptual structure of the database classified by user (DB) constructed | assembled in the memory | storage part of the service server in embodiment of this invention. It is a figure which shows the structural example of the article data delivered to a user in embodiment of this invention. It is a conceptual diagram which shows the basic function of embodiment of this invention. FIG. 8 is a diagram showing a modification (second to fourth operations) of the basic operation (first operation) shown in FIG. 7 together; It is a figure which shows the example of a display of the article data delivered with respect to a certain user in the user terminal of embodiment of this invention. It is a figure which shows the example of a display of the original article received and displayed according to the request | requirement with respect to the article delivered to the user in embodiment of this invention. In an embodiment of the invention, it is a figure showing an example of a recall article displayed when a Recall button is instructed. It is a figure which shows the example of a screen for the recall range setting in embodiment of this invention. It is the flowchart showing an example of the cooperation process between the service server in embodiment of this invention, and a some user terminal. It is a flowchart which shows the modification of the process shown in FIG. It is a flowchart which shows the 1st method of the automatic optimization of the several recall system in embodiment of this invention. It is a flowchart which shows the 2nd method of the automatic optimization of the several recall system in embodiment of this invention. It is a figure for demonstrating the modification of the process of a recall article request | requirement and recall article data delivery in embodiment of this invention.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

  FIG. 1 shows an overall schematic configuration of a data distribution system that provides a service using a communication network in the present embodiment.

  Currently, there are various tools for users to access the Internet 200 as a communication network. In the figure, a PC 100a, a mobile terminal 100b called a smartphone or a tablet, a mobile phone terminal 100c, and a television receiver 100d are shown. The television receiver 100d is shown as a representative home appliance, and various home appliances can be used as the access tool. In this specification, such arbitrary access tools are collectively referred to as a terminal (or user terminal).

  A service server 300 that provides a service according to the present embodiment and a plurality of WEB servers 400 are connected to the Internet 200 as a device that provides a service. The service server 300 constitutes the data distribution apparatus of the present invention. The WEB server 400 includes sites that provide social networking services (SNS) such as information transmission sites, homepages, blogs, Twitter (registered trademark), Facebook (registered trademark), mixi (registered trademark), and the like.

  As a service in the service server 300 according to the present embodiment, a data distribution service that regularly distributes data (typically articles) to user terminals via the Internet is given as an example. For example, the service server 300 voluntarily periodically crawls various sites on the Internet, acquires data of documents, photos, videos, etc. (things that can use text data), and each user (registered user) Data that suits the preference is extracted and given a priority, data having a high priority is selected (or in order of priority), the data is transmitted to the terminal of the user and presented. For this purpose, the service server 300 has the function of a WEB server and the function of sending and receiving electronic mail. In this specification, “user preference” represents user preference, interest, and the like. Information on the Internet can include all kinds of information such as news, posts, advertisements, book information, company information, music information, and the like.

  The applicant of the present application has proposed an information processing method for such an information distribution service as Japanese Patent Application No. 2012-108731. However, the present invention is not based on the information processing method of the prior application.

  The essential function in the present embodiment is that when a user contacts a certain article (data), a similar (or related) article is automatically extracted from the data that the user has contacted in the past, and the user It is a function to present to. Such a function is called a recall function. The recall function is used when a user who contacts an article remembers reading a similar article in the past, but the article cannot be remembered by whom or what article on which site. The corresponding article is detected on behalf of the user and presented to the user.

  Note that “data” that constitutes an article distributed to users in this specification is basically text data of a document consisting of character strings, but it is not available for other media such as still images such as photographs, videos, and music. It may be text data attached to the data.

  FIG. 2 is a block diagram showing various functions of the terminal 100.

  The terminal 100 includes a CPU 101, a storage unit 102, an input unit 104, a display unit 105, and a communication unit 106. Depending on the type of terminal, for example, a voice processing unit 111, a microphone 111a, and a speaker 111b for calling and music player functions may be provided. A terminal such as the television receiver 100d includes a broadcast receiving unit 112. In addition, although not shown, a processing unit unique to each terminal can be provided.

  The CPU 101 is connected to predetermined units, configures a control unit that controls each unit of the terminal 100 by executing various programs stored in the storage unit 102, and implements various functions (means). The storage unit 102 is used as a computer program, an area in which fixed data such as fonts, various setting data, and the like are stored in a nonvolatile manner, a work area used by the CPU 101, and a temporary data storage area. Includes area. Furthermore, the storage unit 102 includes an area for storing various documents and data acquired via the Internet 200 in a nonvolatile manner.

  The input unit 104 is a user interface for the user to input various instructions and data to the terminal 100. Normally, various keys such as a power key, a call key, a numeric keypad, and a cursor operation key can be included. These keys may be hardware keys or may be provided as software. The display unit 105 is a user interface for the terminal 100 to provide display information to the user, and includes a display device such as a liquid crystal display or an organic EL display. The input unit 104 may include a touch panel having a touch input area that overlaps the display screen of the display unit 105. Furthermore, the input unit 104 has an input function using a pointing device such as a mouse, a sensor function that detects the posture and movement of a terminal such as an acceleration sensor, an image sensor function that detects a change in a user's facial expression and facial parts, A function of receiving a voice instruction may be provided. The input of the input unit 104 can include user operation input by these arbitrary functions.

  The communication unit 106 is a means for connecting to the Internet 200, and performs wireless communication with a base station in a third-generation (3G), fourth-generation (4G), etc. mobile phone wireless system via an antenna. And a processing unit for performing a call and data communication with a communication partner via a base station. In addition, as the communication unit 106, any existing communication means such as a wireless LAN or BLUETOOTH (registered trademark) can be used.

  FIG. 3 is a block diagram showing various functions of the service server 300.

  The service server 300 includes a communication unit 310, a display unit 320, an input unit 330, a data processing unit 340, and a storage unit 350 as main functional units.

  The communication unit 310 includes a device that is connected to the Internet 200 and performs data communication, such as a router. The display unit 320 is a user interface for providing display information to maintenance personnel of the service server 300 and includes an arbitrary display device.

  The input unit 330 is a user interface for a maintenance person or the like to input various instructions and data to the service server 300, and is, for example, a keyboard.

  The data processing unit 340 includes a CPU and the like, and is a part that performs various controls of the service server 300 and necessary data processing. In the present embodiment, the data processing unit 340 includes an article acquisition management unit 341, an article distribution processing unit 343, and a user management unit 345.

  The storage unit 350 includes an entire article data storage unit 351, a user-specific article data storage unit 353, and a user management data storage unit 355.

  The article acquisition management unit 341 in the data processing unit 340 is a part that acquires various articles (data) from a site such as the WEB server 400 via the Internet 200, and stores and manages them in the entire article data storage unit 351. is there. The article acquisition management unit 341 also generates a data feature vector from the data for each article stored in the entire article data storage unit 351, selects an article that matches the user's preference for each user, The data is stored in the data storage unit 353.

  The user management unit 345 stores user personal information and user feature vectors in the user management data storage unit 355 as user management data for each individual user. The personal information may include user authentication information (user ID, password, etc.), name, address, nickname, educational background (home school), preference information, etc. when a service is provided for a registered user. Details of the data feature vector and the user feature vector will be described later. Data (for example, user profile data as personal information) may be acquired from the terminal 100. The acquired data is stored in the user management data storage unit 355 together with a user feature vector described later.

The article distribution processing unit 343 is a part that executes processing related to a service that provides articles to the user. As described above, this service projects each user's preference on the Internet service, efficiently discriminates the information (article) necessary (possibly necessary) for the user, and presents it to the user. To do. In particular,
-Matching and recommendation of products and human resources for each user-Prediction of things optimized for users-Filtering of data optimized for users (narrowing)
-It may include searching for data optimized for the user.

  Here, the data feature vector in this specification will be described.

  The data feature vector is a dictionary that associates a weight (a certain real number) with a word appearing in an article (document), and expresses a feature of the article (document). A feature vector can be described as a function of a word, or can be described in the form of a vector if the order of appearing words is fixed as coordinates.

Example 1: A case where a data feature vector is represented by weights as each word included in a document and its frequency distribution.
c = {cat: 0.12, month: 0.03, book: 0.34,….}
This notation means that the word “cat” appears 12% in the document, the error “moon” appears 3%, the term “book” appears 34%, and so on. It can be written in the form of a function such as c (cat) = 0.12, etc., or in the form of a vector such as c = (0.12, 0.03, 0.34,…).

Example 2: A data feature vector is represented by a set of words included in a document.
c = {cat, moon, book,….}
This can be recognized as a dictionary with weights of all 1s, or it can be recognized as a function that gives c (cat) = 1, for example, and 0 for other words in the set. .

Example 3: As the weight in Example 1, not only the frequency in the document but also the document frequency of each word (how many documents appear) may be considered based on a certain document group. TF-IDF (Term Frequency-Inverse Document Frequency) used in the field of information retrieval is one example.

In order to project the preferences of individual users on the Internet service, for example, user feature vectors (preference vectors) are used. In the present specification, the user feature vector represents information specific to a user reflecting the user's preference as a vector. For example, the n-dimensional user feature vector tvec can be expressed as tvec = [a1, a2,..., An]. Here, a1, a2,..., An are n elements of the vector tvec. The user feature vector tvecj for each user j can be expressed as:
tvecj = [aj1, aj2, ..., ajn]

Therefore, the similarity between both feature vectors can be calculated by comparing the data feature vector of each article with the user feature vector. For example, in order to determine the priority (total order) of article data Di for user j, as shown in the following equation, the inner product of user feature vector tvecj and data feature vector cveci of comparison target article data Di is calculated.
(Tvecj · cveci) = Σajm × vim
(M = 1, 2,..., N)
Here, ajm is the value of the mth element of the user feature vector tvecj of the user j, and vim is the value of the mth element of the data feature vector of the data Di of the comparison target article. At this time, the inner product value may be corrected by multiplying the information amount (information entropy) of the data feature vector cvec i. Thereby, the influence which the information amount of data has on the similarity can be reduced.

  Priorities (higher to higher ones) are assigned to the data Di in the order of the size of the inner product. When taking the inner product of the data feature vector and the user feature vector, it is necessary to match the dimensionality of both vectors. In that case, instead of using a virtual maximum dimension number n as described later, it is actually sufficient to assume a dimension number n corresponding to the total number of different words of both vectors.

  Next, priorities are assigned to a plurality of articles according to the calculated similarity. Further, article data for presenting articles according to the assigned priority order is generated. That is, an article with a high priority is preferentially presented to the user. Specifically, only the article with the highest priority is presented, a predetermined number of articles with the highest priority are selected and presented to the user, all articles are presented to the user in the order of priority, etc. Various forms of presentation are conceivable depending on the situation. When presenting, there may be a form in which a plurality of articles are presented step by step according to priority according to a user request.

  Although the present invention is not particularly limited, the technique as presented in the prior application of the present applicant can be used as an example of a specific method for generating the user feature vector. That is, among a plurality of articles presented to the user, in accordance with the operation of the user, a highly interested document in which the user is interested and a low interest document in which the user is not interested are identified, By comparing the word group included in the high interest document and the word group included in the low interest document, the weight value of the word commonly included in both documents is set to “0”, and the word included only in the high interest document is determined. A column of weight values corresponding to a word group in which the weight value is set to a non-zero value is generated as a user feature vector. The similarity is obtained, for example, by comparing the data feature vectors of a plurality of articles to be given priority and user feature vectors, and obtaining the sum of products of weight values of corresponding words in both feature vectors as the similarity. Can do.

  However, the present invention is not particularly limited with respect to specific processing for generating such article distribution data. As a premise of the present invention, it is sufficient that article data is distributed to individual users by some method.

  Since the configuration and function of the WEB server 400 are not directly related to the present invention, description thereof will be omitted. The WEB server 400 receives a request from a terminal (client) such as a content request via the Internet in accordance with http (hyper text transfer protocol), and in response to this, the requested content is sent to the terminal via the Internet. respond.

  FIG. 4 is a block diagram for explaining the overall functions of the data distribution system according to the present embodiment.

  Prior to the use of this service, the user management unit 345 of the service server 300 stores and manages the user identification information and user personal information received from the user terminal 100 in the user management data storage unit 355.

  The article acquisition management unit 341 of the service server 300 periodically (for example, once or multiple times a day, or periodically such as every hour) to specify specific information such as an external information transmission site, an SNS site, and a blog site. The WEB server 400 is accessed, various articles (data) are collected, and stored in the entire article data storage unit 351. Thereafter, article data including a predetermined number of articles to be distributed to each user is created according to the data representing the preferences of each user stored in the user management data storage unit 355 and stored in the user-specific article data storage unit 353. To do. The user-specific article data storage unit 353 sequentially accumulates the created user-specific article data. Further, a mark (or flag) indicating whether or not the user has read is given for each article. The user-specific article data storage unit 353 constitutes means for storing “specific information for specifying past data that the user has contacted in the past” in the present invention.

  The article distribution processing unit 343 reads article data to be distributed to each user from the user-specific article data storage unit 353 periodically (for example, once or several times a day), and the user terminal via a predetermined communication interface To 100. The predetermined communication interface is, for example, a WEB service or electronic mail. In the present embodiment, the article data includes a predetermined number (for example, 30 articles) of articles. As this “predetermined number”, each user may be allowed to designate a desired number.

  When the article reception / display unit 121 of the user terminal 100 receives the article data from the service server 300, the article reception / display unit 121 displays the article data on the display screen of the display unit 105 (FIG. 1). Article data distributed by WEB service or e-mail is described in a markup language document such as HTML (Hyper Text Markup Language). The article reception / display unit 121 includes a mailer program and a browser program. When a so-called link is set in the displayed article data, when the link is instructed (selected) by the user, the article reception / display unit 121 accesses the URL of the link destination, and the content of the access destination Can be further received and displayed.

  When the user who has viewed the displayed article data inputs a recall instruction, the recall instruction detection unit 123 detects the recall instruction and transmits a recall article request to the service server 300. The function of the recall instruction detection unit 123 can be realized by an existing browser program. An article that triggered such a recall instruction is referred to as a “trigger article” in this specification. The “recall article request” includes a trigger article ID (for example, ArticleDATA described later) and a user ID for specifying the trigger article. In addition to the recall article request, predetermined information corresponding to the user operation is transmitted from the user terminal 100 to the service server 300 in a timely manner.

  When the article distribution processing unit 343 receives a recall article request based on a predetermined instruction (that is, a recall instruction) from the user terminal with respect to the distributed article, the article distribution processing unit 343 selects an article similar to the article having received the predetermined instruction. Detection is performed from past articles (past data) in contact with the user in the data storage unit 353. More specifically, by comparing the data feature vector of the article with the data feature vector of each past article of the user, the similarity between the articles to be compared is obtained, and based on the obtained similarity One or a plurality of articles having a high degree of similarity are identified from past articles. Further, the article distribution processing unit 343 returns the article obtained by such detection to the user terminal as recalled article data.

  When receiving the recall article data, the user terminal 100 causes the recall article reception / display unit 125 to display the recall article data on the display screen. The function of the recall article reception / display unit 125 can be realized by an existing browser program.

  FIG. 5 shows a conceptual structure of a database (DB) constructed in the storage unit 350 of the service server 300 in the present embodiment. This DB includes a database for each user shown on the left side of the figure and an entire article database shown on the right side of the figure. The database for each user includes personal information 355 for each user and information for specifying a distribution article (that is, distribution data) or a distribution article for each user stored in the user-specific article data storage unit 353. The user-specific article data storage unit 353 also stores information on article data distributed in the past.

  In this example, the user's personal information includes user identification information (UserID) for identifying individual users, a mail address, and a user feature vector (Taste Vector: tvec).

  Each article stored in the user-specific article data storage unit 353 includes a creation date (Created Date), a delivery date (Delivered Date), an article ID (Article DATA) DATA1, DATA2,..., DATAn, and each article ID. Each data item of a read mark (Selected Mark) provided in the is included. The read mark (Selected Mark) is data indicating whether or not the user has read the article. Here, data indicating unread such as “NO” or “0” is initially set as the read mark, and “Yes” or “ The data is changed to data indicating the read such as “1”. Other data items are set when article data is created.

  The whole article database includes a plurality of article data. Each article data includes the article ID (Article DATA), title (Title), text (Text), article URL (Article URL), feed URL (Feed URL), thumbnail URL (Thumbnail URL), feature vector (Characteristic) Vector: cvec), accompanying information (Info), etc.

  The text (Text) of each article data is composed of a character string representing a part of the article document or the whole sentence or a summary. Article URL (Article URL) represents the address of the distribution source of the article. A feed URL is a URL for news distribution at an information transmission site (such as a newspaper company) that provides the article. A thumbnail URL (Thumbnail URL) represents an address of an image related to the article. The thumbnail URL may not exist depending on the article. A feature vector (Characteristic Vector: cvec) is a data feature vector of the article. The accompanying information (Info) is arbitrary information related to this article, for example, information indicating that the article is an article taken from the external storage.

  Whether or not the user has read the article can be determined, for example, that the user has read the article typically when the user performs an operation to display the full text on the distributed article. Articles read by the user are included in articles that the user “contacts”. However, the articles touched by the user are not limited to the articles read by the user, and articles touched by various buttons to be described later can be included in the “touched articles”.

  FIG. 6 shows an example of the structure of article data distributed to the user.

  The distribution article data includes a distribution date and article data of a predetermined number of articles. Each article data includes a title, text, URL, thumbnail URL, and various buttons. The buttons here are, for example, data elements corresponding to images and functions corresponding to operation buttons displayed on the display screen. In this example, Like! Examples of buttons, Share buttons, and Comment buttons (see FIG. 9 described later) are shown. A link is set to each of these elements (items) as necessary. When the element is instructed (selected) by the user, the link is moved to the link destination or a specific function (process) associated with the link is activated. As a specific example, such a structure is converted into HTML data and distributed.

  Hereinafter, a schematic configuration of a service in the present embodiment will be described, and then a specific processing example will be described.

  FIG. 7 is a conceptual diagram showing the basic functions of the present embodiment. When the user A issues a recall instruction from the terminal to an article distributed to the user A, from the article DB that stores the past articles of the user A in which the articles that the user A has contacted in the past are accumulated. Articles similar to (relevant to) the trigger article are automatically searched. An article touched in the past is, for example, an article whose read mark indicates read. Therefore, the similarity between the articles is calculated based on the data feature vector of the trigger article and the data feature vector of the article data (past article data) of the past article of the user A with the read mark. A predetermined number of articles are extracted from the past articles having a high degree of similarity and presented to the user A (terminal).

  As a specific example of obtaining the similarity between articles, here, the data feature vector of the article is (w1, v1), (w2, v2), ... (wn, vn) as shown in Example 1 above. A case where the word (w) and its weight (v) are represented in a paired format will be described as an example.

More specifically, as the similarity between articles, as an example, the correlation coefficient cvec0 · cveci / () between the data feature vector cveci of the trigger article and the data feature vector cveci of the data Di of the past article is represented by the following equation. | cvec0 || cveci |) is calculated.
cvec0 · cveci / (| cvec0 || cveci |) = Σam × vim / (| cvec0 || cveci |)
(M = 1, 2,..., N)
Here, am is the value of the mth element of the data feature vector cvec0 of the trigger article, and vim is the value of the mth element of the data feature vector cveci of the data Di.

  Priorities (higher to higher ones) are assigned to the data Di in the order of the correlation coefficient values. When taking the inner product of data feature vectors, it is necessary to match the dimensionality of both vectors. In that case, instead of using a virtual maximum dimension number n as described later, it is actually sufficient to assume a dimension number n corresponding to the total number of different words of both vectors.

  Next, priorities are assigned to a plurality of past articles (data) according to the calculated similarity. Furthermore, an article is presented (or processed) according to the assigned priority order. That is, an article with a high priority is preferentially presented to the user. Specifically, only the article with the highest priority is presented, a predetermined number of articles with the highest priority are selected and presented to the user, all articles are presented to the user in the order of priority, etc. Various forms of presentation are conceivable depending on the situation. When presenting, there may be a form in which a plurality of articles are presented step by step according to priority according to a user request.

  FIG. 8 is a diagram showing a modification (second operation) in addition to the basic operation (first operation) shown in FIG. The first operation is as described above.

  The second operation shown in FIG. 8 is to convert (extend) the data feature vector of the trigger article into an associative vector, and then apply the associative vector to the article DB of user A in the same manner as described above, thereby The upper predetermined number of articles having data feature vectors that are similar to each other are extracted. A specific example of the association vector will be described later.

  FIG. 9 shows a display example of article data distributed to a user on the user terminal. In this example, it is assumed that article data is distributed regularly (for example, once a day) to each registered user of the data distribution service.

  The number of articles delivered to the user at a time may be variably set on a setting screen (not shown).

  In the distributed article data display screen 500, the distribution date 501 is displayed at the top, and a plurality of article areas 510 are arranged below it.

  In each article area 510, a title 511, text 512, various buttons 513, and thumbnail images 514 are arranged in a predetermined positional relationship. For example, a link to a URL indicating the original article (content) of this article is set in the title 511. Text 512 is a portion or summary of the original article. When the user instructs the title 511, an original article is requested from the WEB server, and the received original article is displayed as a display screen 610 in FIG. 10 instead of (or overlaid) the display screen 500.

  Here, it is assumed that the content displayed on the display screen 610 is acquired from the original article site. Accordingly, various buttons 513 as shown in FIG. 9 are not displayed on the display screen 610 of FIG. If the same various buttons 513 are included in the original screen, they are also displayed on the display screen 610.

  Although it is assumed that the link to the original article is set to the title 511, it may be set to the text 512 instead of (or in addition to) the title 511. Alternatively, a character string such as “display full text” or “view continuation” may be additionally displayed, and a link may be set to this character string.

  In addition, there may be a mode in which the information of the entire text is originally included in the distribution article data and displayed on the display screen 500 in a state where a part of the information is hidden. In this case, the distribution article data can be configured such that the link to the original article is not set and the full-text display is switched when an instruction is given to the title 511 or the like. In such a case, various buttons 513 can be prepared on the full text display screen.

  Various buttons 513 correspond to the example shown in FIG. Like! The button is a button that the user instructs so that the user who reads this article expresses approval. The Share button is a button that the user has instructed to publish on the SNS to which the user who has read this article shares to share with other users. The Comment button is a button for the user who has read this article to instruct in order to write a comment. The “Recall” button is a button for the user who has read this article to instruct the user to activate the recall function according to the present invention. These various buttons 513 can be said to be a user operation representing an intention display in which the user is interested in the article. Such buttons for user operations are not limited to those illustrated.

  When a user issues a recall instruction to the Recall button for any article on the display screen 500, one or a plurality of articles similar to (or related to) the article from past articles that the user has contacted Articles are extracted and displayed as recall articles as shown in the display screen 700 of FIG. Although the display area 710 shown here is shown as corresponding to the article area 510 shown in FIG. 9, it may have a different format. For example, the recall article can be displayed in various formats such as only the title, title and body, title and body and image. Also, the display position of the recalled article can be considered variously such as taking out under the article area on the display screen 500, putting it up, putting it to the right, putting it over.

  An example of a screen for setting a recall range will be described with reference to FIG.

  In the above description, when the recall instruction is issued by the user, the past data that is the target of detecting the recall article is assumed to be an article that the user has contacted in the past. On the other hand, the following expansion is possible.

  As described above, the “article contacted by the user in the past” is typically an article read by the user in the past, but may include articles in which various button operations have been performed. Furthermore, the article may simply include a recommended article. In the present embodiment, “recommended articles” are all articles distributed to the user as displayed on the display screen 500 as described above.

  On the other hand, in recent years, an external storage service has been used for a user to store a large amount of data. With external storage services, it is difficult to extract necessary information other than keyword search and folder organization. Therefore, in addition to the articles presented (recommended) by the article distribution service, articles acquired from an arbitrary WEB server by using an external storage service or the like are included in the “past data”. For this purpose, articles in the external storage are stored in the entire article data storage unit 351 and referred to based on the contents of the user-specific article data storage unit 353. A plurality of external storage services that are used may be individually specified.

  Furthermore, in the above description, the “past data (past article)” is limited to the past data (past article) of the user who issued the recall instruction. On the other hand, “past data” can be extended to past data of users other than the user who issued the recall instruction. As “another user”, a specific user may be designated by an e-mail address or a user ID. Also, it is possible to designate “users with the same profile” as themselves, designate “users with similar preferences”, and designate “all users” other than themselves.

  Such various options can be arbitrarily set (save) by the user on the recall range setting screen shown in FIG. This recall range setting data can be stored in the user management data storage unit 355 of the service server 300, for example.

  In order to avoid privacy problems, such other users are preferably limited to users who have consented to publish their past data to others. The “others” may be all users other than themselves, or may be set to an option such as opening only to a specified user or opening to a user other than the specified user.

  FIG. 13 is a flowchart showing an example of cooperation processing between a service server and a plurality of user terminals in the present embodiment. Although two user terminals, a user A terminal and a user B terminal, will be described here, a larger number of user terminals may actually be involved.

  Prior to article distribution to the user, the service server 300 periodically sends a predetermined site on the Internet (for example, the site represented by the above-described feed URL) via the communication unit 310 by the article acquisition management unit 341. Crawl and collect articles. Collected articles are accumulated in the entire article data storage unit 351. Thereafter, at a predetermined time, a predetermined number (for example, 30 articles) of each user is extracted using the user feature vector of each user, and is stored in the user-specific article data storage unit 353 (S211). The service server sequentially distributes the latest article data created for the user to each user terminal via the communication unit 310 by the article distribution processing unit 343 via the communication unit 310 on a regular basis or at any time or in response to a user request. (S212). You may perform the extraction of the article | item for every user in the case of the delivery to the said user.

  When the terminal of user A receives the distribution article data addressed to itself from the service server via the communication unit 106 (S110, Yes), the display unit 105 displays the article data as shown in the display screen 500 of FIG. Is displayed (S111). A user operation from the input unit 104 is accepted for the displayed article (S112). Although not specifically described, the terminal of the user A makes a corresponding response depending on the contents of the user operation. Operation information is also notified to the service server for updating the read mark or the like. When the various buttons described above are instructed, the terminal of user A executes corresponding processing. In particular, when the user operation is “recall instruction”, the terminal of user A requests a recall article corresponding to the trigger article to the service server (S113).

  In response to the recall article request, the service server refers to the user's past data in the user-specific article data storage unit 353, extracts the recall article corresponding to the trigger article, and uses the terminal of user A as the recall article data. (S213).

  The terminal of user A that has received the recall article data displays the recall article on the display screen (S114). User operations are also accepted for the display of recalled articles. When an instruction to end the article data display is given (S115, Yes), the process returns to the first step S110. When an instruction to return to article data display is made (S116, Yes), the process returns to step S111. Otherwise, the process returns to step S112.

  The same processing as that of the user A terminal is performed for the user B terminal (S120 to S126, S214).

  FIG. 14 shows a modification of the process shown in FIG. Steps similar to those shown in FIG. 13 are denoted by the same reference numerals, and redundant description is omitted. In the process of FIG. 13, the case where the recall article extraction process is executed in real time in response to the recall article request is shown. On the other hand, in the process of FIG. 14, instead of extracting a recall article in real time according to a recall instruction, a recall article is also extracted for each article to be distributed in advance when creating distribution article data. deep.

  That is, when the service server 300 extracts a predetermined number of articles (for example, 30 articles) for each user by using the user feature vector of each user by the article acquisition management unit 341 prior to article distribution to the user ( S221), recall articles for each article are also extracted for each user (S222). The extracted recall article (at least information for specifying the article) is stored in the user-specific article data storage unit 353 in association with the user and the article. Further, the service server sequentially distributes the article data created for the user to each user terminal on a regular basis or at any time or in response to a user request (S223). The article data distributed to each user is embedded in such a form that a recall article created for the user is initially hidden on the terminal and displayed in response to the user's recall instruction.

  Since the processing of the user terminal is basically the same as the case of FIG. 13, only the processing characteristic of the processing of FIG. 14 will be described below. At the stage of article data display (S111, S121) on the user terminal, the recall article is hidden and is not displayed. When the user terminal receives a recall instruction from the user (S113, S123, Yes), the user terminal reads the recall article data stored inside (the storage unit 102) together with the received article data (S113a, S123a). It is displayed (S114, S124).

  In the method of FIG. 13, recall article extraction is executed by real-time processing. Therefore, if this processing takes time, the user is kept waiting. However, the method of FIG. 14 does not have such an adverse effect. In contrast, in the method of FIG. 14, recall articles are prepared for all distribution articles that are not necessarily instructed to be recalled in the future, so the processing load on the service server side increases, and the data volume of the distribution articles However, the method of FIG. 13 does not have such an adverse effect.

  In addition, when recall articles are extracted in advance, it is assumed that they are embedded in the article data to be distributed, but can be stored on the service server side without being embedded in the article data. In this case, there is no need to wait for the user when a recall instruction is given, and there is no need to transmit extra data when distributing an article.

  Next, an automatic optimization method using a plurality of recall methods will be described. As described in FIG. 8, a plurality of calculation algorithms can be used as a method for extracting a recall article triggered by an article. It is also conceivable that the system automatically selects these calculation algorithms and tunes them according to the user's preference.

  FIG. 15 is a flowchart showing a first method of automatic optimization of such a plurality of recall methods. In this method, a plurality of recall methods are used in combination to create recall article data at one time. Here, an example will be described in which a predetermined number of articles (30 articles in this example) are extracted by combining the first, second, and third recall methods. For example, the first recall method is the first recall method shown in FIG. 8, the second recall method is the second recall method shown in FIG. 8, and the third recall method is shown in FIG. In the second recall method, the associative vector is made different.

  When generating the recall article data, the service server first extracts the top n1 articles by the first recall method (S231). Next, the top n2 articles are extracted by the second recall method (S232). Further, the top n3 articles are extracted by the third recall method (S233). Here, the initial values of n1, n2, and n3 are the same (here, 10). Therefore, after waiting for a recall instruction from the user terminal (S234), recall article data including these extracted articles # 1 to # 30 is transmitted to the user terminal (S235).

  The user terminal receives the recall article data and displays the articles # 1 to # 30 (S131). Thereafter, if there is a selection operation for any recall article (S132, Yes), the selected article information is notified to the service server.

  Here, the “selection operation” is, for example, an instruction operation for displaying the above-described full sentence or various button operations described above for the article. That such a selection operation has been performed means that the user has expressed some interest in the recall article. That is, it means that the recall method from which the recall article is extracted is appropriate for the user.

  Further, the “selected article information” here is information for specifying the recall article included in the recall article data, for example, and is a numerical value indicating the number of the recall article in the recall article data. In accordance with this information, the service server knows by which recall method the article to which the selection operation is directed was extracted.

  In response to the selected article information from the user terminal, the service server updates the presentation order of the extracted articles and / or the number of extracted articles based on the recall method determined to be appropriate (S236). The “presentation order of extracted articles by each recall method” is the order of articles in the recall article data, and in this example, is the execution order of the first, second, and third recall methods (presentation order in execution order). ) The first, second, and third recall schemes are set in order from the recall scheme with the highest priority. “Updating the number of articles extracted by the recall method” is the number of articles extracted by the recall method. For example, changing the values of the number of shared articles n1, n2, and n3 without changing the total number of extracted articles of the articles of the first, second, and third recall methods.

  By using such an automatic optimization method, it is possible to increase the frequency (or priority) of using a recall method suitable for each user.

  FIG. 16 is a flowchart showing a second method of automatic optimization of a plurality of recall methods. In the first method of FIG. 15, a plurality of recall methods are used at the time of creating the recall article data at one time. However, in the second method, at the time of creating the recall article data once, only a single recall method is used and created. The recall method is switched every hour.

  In response to the recall instruction (S241), when generating the article data, the service server first extracts the top 30 articles # 1 to # 30 by the n-th recall method to the user terminal as recalled article data. Transmit (S242). (The initial value of the variable n is 1, for example.)

  The user terminal receives the recall article data and displays the articles # 1 to # 30 (S141). After that, if there is a selection operation for any recall article (S142, Yes), the selected article information is notified to the service server.

  Upon receiving the selected article information notification from the user terminal, the service server accumulates the relationship between the selected article information and the recall method, for example, in the storage unit 350 (S243).

  Therefore, it is checked whether the user applicability of the current recall method is significant (S244). Whether or not it is significant can be determined by determination based on the presence or absence of a recall article selection operation, or by a threshold determination of the selection operation. The threshold determination of the selection operation is, for example, a determination of whether the ratio of the number of selections to the number of recall article distributions is a predetermined threshold or more. If not significant, the variable n is incremented (S245), and the process returns to step S241. The variable n is cyclically incremented. For example, in this example, n changes as 1, 2, 3, 1, 2, 3,.

  If found to be significant (S244, Yes), the process returns to step S241 without incrementing n in order to maintain the current recall method. Such n-value processing corresponds to increasing the frequency of use (or priority) of the recall method with significant user applicability.

  Such an automatic optimization method can increase the frequency (specific gravity) at which a recall method suitable for each user is used.

  FIG. 17 illustrates a modification of the recall article request and recall article data distribution process. In the processing described above, it is assumed that the recall instruction for the same article is at most once. However, when a recall article is presented in response to the user giving a recall instruction once, the article (or any of them) may not be an article requested by the user. Therefore, the user is allowed to give further recall instructions to the same trigger article, and the recall article is extracted by a recall method different from the previous one in response to the recall instruction request. Thereby, when the article to be requested is not extracted due to the nonconformity of the recall method, the possibility that the requested article is extracted by a further recall instruction increases.

  FIG. 17 shows processing steps obtained by modifying the main part of FIG. 13 with respect to such a step-by-step recall instruction. When there is a recall instruction from the terminal of user A (S113, Yes), the recall instruction is confirmed by the (p) th recall method for the service server after confirming what the recall instruction is (p) th recall instruction for the article. Is requested (S113a). In response to the recall article request (p), the service server extracts a recall article corresponding to the trigger article by the p-th recall method and distributes it to the terminal of user A (S213a). At this time, preferably, articles that overlap with the recall articles that have already been transmitted are excluded. The terminal of user A receives and displays this recall article (S114).

  The upper limit of the numerical value p is set according to the number of recall methods. Further, a special button such as a recallmore button (not shown) may be prepared for “further recall instruction”, or the number of times the original recall button is pressed may be counted for identification. Furthermore, a separate menu may be used. As the input means, any of the above-described user operation input means can be used.

  Next, the “association vector” described above will be described. As described above, the associative vector is an extension (change) of the data feature vector of the trigger article, and an outline and a specific example of the change method will be described below.

  In this embodiment, the associative vector (rvec) is obtained by applying a user feature vector of user A, some weighted word graph expressing user A's preference, etc. based on the data feature vector of the trigger article. This is a vector in which features are expanded, that is, a vector that represents features that the user A associates with a trigger article. In order to obtain such an associative vector, for example, the obtained related word is added to the data feature vector cvec using a dictionary of related words that associates a related word with a certain word.

For example, if the data feature vector cvec = {(curry 0.3), (spicy 0.2), (restaurant 0.1),…}
Association vector rvec = {(Curry 0.3), (India 0.2), (Spicy 0.2), (Restaurant 0.1), (Indian cuisine 0.1),…}
Is obtained.

  Hereinafter, a specific example in which an association vector is obtained using the relationship matrix M will be described. Hereinafter, for convenience, the feature vector is represented by c, and the association vector is represented by r.

  The relationship matrix M is a dictionary that associates weights (real numbers) with two word pairs, and expresses the strength of the relationship between words. In addition to being able to write as a function for a pair of words, the word “cat” is the first coordinate, the word “moon” is the second coordinate, etc. Can express.

Example 1: For a group of documents (articles) selected by a user, a set of pairs of words that are simultaneously included in one document (that is, “co-occurring”).
M = {(Cat, Dog), (Cat, Moon), (Month, Book),… ..}
This is a dictionary in which a value of 1 is associated with word pairs included in one document and a value of 0 is associated with other word pairs. It can also be written in the form of a function such that the relation matrix M ((cat, dog)) = 1.

Example 2: Consider the feature vector of each document (article) group selected by the user. For word pairs included in one document (that is, one feature vector) at the same time, the weight product in each feature vector is averaged over all the feature vectors included at the same time.
M = {(cat, dog): m,….}, M = M ((cat, dog)) = (1 / N) Σ c (cat) c (dog), etc.
(Here, N is the number of feature vectors including “cat” and “dog” at the same time. The summation Σ covers the entire feature vector c.)

Using the relationship matrix M, the associative vector r can be obtained as follows. The associative vector r is given by a functional F with respect to a given feature vector c and relation matrix M.
r (w) = F (M, c) (w)
Is a dictionary that associates the weight r (w) with the word w. The associative vector r obtained by this equation can include information on words that did not exist in the feature vector c.

  In this specification, “association” means that an associative vector r = F (M, c) is created for a given feature vector c and relation matrix M 1, and a feature vector “similar” to the associative vector r This is an operation to select an article (document) having a from an article group (document group).

  As a result, it is possible to search for a target based on a wider “association” that takes into account the relationship expressed by the association matrix, rather than simply searching for an object close to the original feature vector c 1.

  The “similarity” between vectors can be obtained from the correlation coefficient (c · r) / (| c || r |) or Euclidean distance | c-r |.

  Examples of association vectors include the following. The associative vector can vary depending on how the relationship matrix M and functional F are selected. (It also depends on the definition of feature vectors).

Example 1: F (M, c) = M c, that is, the product in the ordinary sense of the relation matrix M and the feature vector c. Specifically,
r (w) = Σ M (w, w ') c (w'). (The summation Σ covers all words w ')

  Example 2: In the above example, instead of the sum, take the maximum value for all w '.

Example 3: F (M, c) = M ² c, that is, the product of the square of the relation matrix M and the feature vector c.

More specifically,
r (w) = Σ (Σ M (w, w ') M (w', w``)) c (w '').
(The summation Σ is for all words w 'on the inside and for all words w''on the outside).

  In Example 1 and Example 2, a word that was not included in the original feature vector is included if it is related to the word in the relation matrix (for example, a word that appeared in the same document at the same time). In the same meaning, the original word is added to the word related to the related word.

Example 4: (Resolvent)
For the relation matrix M and small parameter ε, the following inverse matrix can be calculated,
R (M) = (1-εM) ^-1
And calculate r = R (M) c.
R (M) = 1 + εM + ε ² M ² + ε ³ M ³ +…
Therefore, this is a combination of all of the relation relations, relation relation relations (with a weight that decreases as the distance increases).

  A simple specific example of associative vector calculation is given below.

Given feature vector (the data feature vector of the article the user has just selected)
c = {cat: 0.3, month: 0.5, sake: 0.2}
And Also, given feature vector group (data feature vector group of articles previously selected by user)
c1 = {Cat: 0.6, Dog: 0.4},
c2 = {Month: 0.1, Book: 0.7, Movie: 0.2}
And

From cat 1 to coordinate 6 in the order of cat, moon, drink, dog, book, movie,
c = (0.3, 0.5, 0.2, 0, 0, 0),
c1 = (0.6, 0, 0, 0.4, 0, 0),
c2 = (0, 0.1, 0, 0, 0.7, 0.2)
And vector type.

In addition, as a relational matrix created from c, c1, and c2 (depending on whether c is also used or not), a 6 × 6 matrix with 1 for words co-occurring in the same vector and 0 otherwise make. That means
(Cat, Moon), (Cat, Liquor), (Month, Liquor); (Cat, Dog); (Month, Book), (Month, Movie), (Book, Movie) ), (1, 3), (2, 3); (1, 4); (2, 5), (2, 6), (5, 6) coordinates and the coordinates of the rows and columns reversed, And let M be a matrix with only 1 diagonal elements and 0 otherwise. Here, (p, q) coordinates mean p rows and q columns.

すなわち、r = (1, 1, 1, 0.3, 0.5, 0.5) となり、c に含まれない「犬」、「本」、「映画」もある重みをもって連想される。 When the associative vector r is calculated by r = Mc, the following equation (1) is obtained.

That is, r = (1, 1, 1, 0.3, 0.5, 0.5), and “dog”, “book”, and “movie” not included in c are also associated with certain weights.

  The preferred embodiments of the present invention have been described above, but various modifications and changes other than those mentioned above can be made. For example, the language of the document has been described only for Japanese, but other languages may be used. In the data distribution system, the user terminal and the server are connected via the Internet. However, a local communication network such as a LAN may be used instead of the Internet.

  A computer program for realizing the functions described in the above embodiments by a computer and a recording medium storing the program in a computer-readable manner are also included in the present invention. “Recording media” for supplying the program include, for example, magnetic storage media (flexible disks, hard disks, magnetic tapes, etc.), optical disks (magneto-optical disks such as MO and PD, CDs, DVDs, etc.), semiconductor storage, etc. Can be mentioned.

DESCRIPTION OF SYMBOLS 100 ... User terminal 100c ... Mobile phone terminal 100d ... Television receiver 102 ... Memory | storage part 104 ... Input part 105 ... Display part 106 ... Communication part 111 ... Audio | voice processing part 111a ... Microphone 111b ... Speaker 112 ... Broadcast receiving part 121 ... Article reception Display unit 123 ... Recall instruction detection unit 125 ... Recall article reception / display unit 200 ... Internet 300 ... Service server 310 ... Communication unit 320 ... Display unit 330 ... Input unit 340 ... Data processing unit 341 ... Article acquisition management unit 343 ... Article Distribution processing unit 345 ... User management unit 350 ... Storage unit 351 ... Whole article data storage unit 353 ... User-specific article data storage unit 355 ... User management data storage unit 400 ... WEB server 500 ... Display screen 501 ... Distribution date 510 ... Article area 511 ... Title 512 ... Text 513 ... Various buttons 5 4 ... thumbnail image 610 ... display screen 700 ... display screen 710 ... display area

Claims (13)

A data distribution method for distributing data to a user terminal,
Delivering data to a user terminal;
Storing specific information for specifying past data that the user of the user terminal has contacted in the past;
In response to a predetermined instruction from the user terminal with respect to the distributed data, a step of detecting data similar to the data that has received the predetermined instruction among past data that the user has contacted;
Transmitting the detected data to the user terminal. For data distributed to the user terminal, further comprising a step of generating a data feature vector representing a feature of each data,
In the detecting step,
By comparing the data feature vector of the data with the data feature vector of the past data, the similarity between the data to be compared is obtained, and the similarity is high among the past data based on the obtained similarity The data distribution method according to claim 1, wherein one or a plurality of data is specified.   The data distribution according to claim 2, wherein in the step of obtaining the similarity, a data feature vector of the distributed data is compared with a data feature vector of the past data, and a correlation coefficient in both feature vectors is obtained as the similarity. Method.   The data distribution method according to claim 2, further comprising the step of converting the data feature vector into an associative vector obtained by extending the feature, and in the detecting step, the associative vector is compared with a data feature vector of the past data. .   The data distribution method according to claim 1, wherein the detecting step is executed in real time according to the predetermined instruction.   5. The data distribution method according to claim 1, wherein the detecting step is executed in advance for each data to be distributed when the data is distributed.   In the distribution step, the past data obtained by the detecting step executed in advance is initially hidden in the terminal, and is embedded in the distribution data so as to be displayed according to the predetermined instruction. The data delivery method according to any one of claims 1 to 6.   The “article contacted by the user in the past” is at least one of an article read by the user in the past, an article that has been operated by a user representing an intention indicating that the user is interested, and an article that has been received in the past. The data distribution method according to claim 1.   The data distribution method according to claim 1, further comprising a step of managing data stored by a user using an external storage service as data that the user has contacted in the past. After the detected data is transmitted to the user terminal, when a predetermined instruction is received again for the same data as the data that has received the predetermined instruction from the user terminal, Changing the detection algorithm, and again detecting data similar to the data that has received the predetermined instruction;
The data distribution method according to claim 1, further comprising: transmitting the detected data to the user terminal again. A data distribution device for distributing data to a user terminal,
Distribution data generation means for generating data to be distributed to the user terminal;
Past data storage means for storing specific information for specifying past data that the user of the user terminal has contacted in the past;
In response to a predetermined instruction from the user terminal with respect to the distributed data, data detection means for detecting data similar to the data that has received the predetermined instruction among past data that the user has contacted;
A data distribution apparatus comprising: transmission means for transmitting the detected data to the user terminal. A computer program for causing a computer to execute a data distribution method for distributing data to a user terminal,
Delivering data to a user terminal;
Storing specific information for specifying past data that the user of the user terminal has contacted in the past;
In response to a predetermined instruction from the user terminal with respect to the distributed data, a step of detecting data similar to the data that has received the predetermined instruction among past data that the user has contacted;
A computer program for causing a computer to execute the step of transmitting the detected data to the user terminal.   The recording medium which recorded the computer program of Claim 12 so that computer reading was possible.

Images