JP2008015694A - Document taste learning system, method, and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To reduce or dispense with intentionally carried out operation (work) required from a user for learning a document taste of the user. <P>SOLUTION: A document taste learning system counts bookmark registration or file storage of a document as one condition for determining that the document is attractive to the user. Another document taste learning system counts a longer staying time of the user in browsing of a document in comparison with a threshold value as one condition for determining that the document is attractive to the user. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a document preference learning system, method, and program, and can be applied to, for example, efficiently searching for a document desired by a user.

  In recent years, with the spread of computers and the Internet, many Web pages have been generated. As a result, since the number of Web pages has become enormous, there is an increasing demand for efficiently searching for Web pages requested by users. As a solution, a technique such as Web mining using Bayesian estimation is being studied (see Non-Patent Document 1 and Non-Patent Document 2).

In this conventional method, by learning the characteristics of a user's favorite web page (web page of interest), the web page that the user is similarly interested in (same as the search information such as a keyword) is not considered. With the goal of realizing a system that collects hobby-preferred web pages), a search method including a method for learning favorite web pages has been proposed.
Tamano Amano, Hidenori Nakazato, “Web mining using Bayesian inference”, Proceedings of Society Conference of IEICE Society Society, No. B-19-9, p. 404, September 2004 Tamano Amano, Hidenori Nakazato, Takashi Nakamura, “Web Mining Using Bayesian Estimation”, IEICE Technical Report, IEICE, March 15, 2005

  However, according to the conventional method, the user needs to manually collect and classify pages that are not related to a favorite page. That is, in order to learn the user's document preferences (documents of interest), the user himself / herself needs to select the likes (interests) and dislikes (no interests) for the information on the Web page (URL). And it takes a lot of work in advance learning. For example, a “like” icon or a “dislike” icon is displayed on a part of a web browser screen displaying a web page, and the user learns his / her preference by clicking on either icon.

  In order to improve the accuracy of the search, it is preferable that there are many Web pages to be learned, and if the number of pages to be learned is increased, the burden of manual work by the user becomes considerably large.

  The present invention has been made in view of the above problems, and a document preference learning system, method, and program capable of reducing or eliminating an operation (work) that a user intentionally executes in order to learn a user's document preference. The purpose is to provide.

  In the document preference learning system according to the first aspect of the present invention, a feature amount obtaining unit for obtaining a feature amount of an input document and an act of storing information related to the input document for future use and use of the input document are performed. An interest discriminating means for discriminating whether or not the user is interested in the input document, and a feature amount of the input document as one condition for discriminating whether or not the act of saving is performed as a document in which the user is interested. And a document feature amount storage means for storing the information according to whether the user is interested in the document.

  A document preference learning system according to a second aspect of the present invention is characterized in that a feature amount acquisition unit that obtains a feature amount of an input document, and that a user is interested in a document whose time during which the input document is browsed exceeds a threshold. As one condition, there is an interest determination unit that determines whether or not the user is interested in the input document, and a document feature amount storage unit that stores the feature amount of the input document according to whether the user is interested in the document. It is characterized by that.

  A document preference learning method according to a third aspect of the present invention includes a feature amount acquisition unit, an interest determination unit, and a document feature amount storage unit. The feature amount acquisition unit obtains a feature amount of an input document, and the interest determination The means determines whether or not an act of storing the information related to the input document for the future use and use of the input document is performed, and the user is interested in the fact that the act of storing is performed. As one condition for determining, whether or not the user is interested in the input document is determined, and the document feature amount storage means distributes and stores the feature amount of the input document according to whether the user is interested in the document. And

  A document preference learning method according to a fourth aspect of the present invention includes a feature amount acquisition unit, an interest determination unit, and a document feature amount storage unit. The feature amount acquisition unit obtains a feature amount of an input document, and performs the interest determination. The means determines whether or not the user is interested in the input document as one condition for determining that the user's browsing time of the input document exceeds a threshold as a document of interest to the user, and stores the document feature value storage. The means is characterized in that the feature quantity of the input document is sorted and stored according to whether the user is interested in the document.

  In the document preference learning program of the fifth aspect of the present invention, the computer stores feature quantity acquisition means for obtaining the feature quantity of the input document, and information related to the input document for future use and use of the input document. An interest discriminating means for discriminating whether or not the user is interested in the input document as one condition for discriminating whether or not an act has been performed, and discriminating that the act of saving has been performed as a document of interest to the user, and the input document The feature amount is described so as to function as a document feature amount storage unit that stores the feature amount according to whether the user is interested in the document.

  In the document preference learning program of the sixth aspect of the present invention, the user is interested in the feature amount obtaining means for obtaining the feature amount of the input document and the time that the user is browsing the input document exceeds the threshold. As one condition for discriminating a document, an interest discriminating unit that discriminates whether or not the user is interested in the input document, and a document feature amount storage unit that stores the feature amount of the input document according to whether the user is interested in the document. It is described to function as.

  ADVANTAGE OF THE INVENTION According to this invention, in order to learn a user's document preference, the operation | movement (work | work) which a user intends to perform can be reduced or eliminated.

(A) First Embodiment Hereinafter, a first embodiment in which a document preference learning system, method and program according to the present invention are applied to a Web page browsing system, method and program will be described in detail with reference to the drawings.

(A-1) Configuration of First Embodiment FIG. 1 is a block diagram illustrating a functional configuration of a Web page browsing system according to the first embodiment. The Web page browsing system of the first embodiment installs the Web page browsing program (including fixed data; so-called Web browser) of the first embodiment on an information processing apparatus such as a personal computer, a mobile phone terminal, or a PDA. Although it is constructed by this, it can be functionally represented in FIG.

  In FIG. 1, a Web page browsing system 1 according to the first embodiment includes an input unit 10, a display unit 11, a communication unit 12, a browsing control unit 13, a page feature value forming unit 14, a page feature value storage unit 15, and a preference document. A determination unit 16 and a bookmark storage unit 17 are included.

  The input unit 10 corresponds to hardware such as a keyboard and a mouse, and instructs to start browsing a Web page, searches for a user-preferred document, and instructs registration to a bookmark (favorite URL). It is something to do.

  The display unit 11 corresponds to a display in terms of hardware, and displays a web page for browsing or displays various operation icons.

  The communication unit 12 corresponds to a communication circuit, communication software, or the like, and executes communication for retrieving the Web page instructed from the browsing control unit 13 from the Internet 20.

  The browsing control unit 13 controls various operations related to browsing of Web pages according to a signal given from the input unit 10 and a processing stage at that time.

  The page feature amount forming unit 14 forms the feature amount of the Web page instructed from the browsing control unit 13. The feature amount of the Web page is formed when learning the user's document preference or when determining whether or not the searched Web page matches the user's document preference.

  The page feature amount storage unit 15 stores a page feature amount reflecting the user's document preference. The page feature amount storage unit 15 stores a feature amount of a Web page that the user likes (with interest) and a feature amount of a Web page that the user dislikes (with no interest) in distinction. The page feature quantity storage unit 15 may be entirely or partially configured on an auxiliary storage device.

  The preference document determination unit 16 determines whether the Web page instructed from the browsing control unit 13 matches the user's document preference. At the time of determination, the preference document determination unit 16 acquires the feature amount of the Web page to be used for determination from the page feature amount formation unit 14 and also considers the storage content of the page feature amount storage unit 15.

  The bookmark storage unit 17 stores a URL specified by the user as a bookmark under the control of the browsing control unit 13.

(A-2) Operation of First Embodiment Next, an operation (Web page browsing method) in the Web page browsing system of the first embodiment will be described.

  First, the operation when learning the user's document preference will be described with reference to the flowcharts of FIGS. FIG. 2 shows the flow of the overall operation when learning the user's document preference for a certain Web page, and FIG. 3 shows the flow of the page feature value creation operation. In the middle of the processing of FIG. 2 and FIG. 3, other processing may be interrupted by key input or the like, but in FIG. 2 and FIG. 3, such interrupt processing is omitted.

  In the Web page browsing system of the first embodiment, when a new Web page is acquired from the Internet 20 for use by the user, the browsing control unit 13 starts the process shown in FIG. The Web page that has been displayed is displayed (viewed) (step 101). This display processing may include a case where the Web page already acquired from the Internet 20 and stored in the cache memory is displayed by displaying another Web page. Conversely, such a case is excluded. You may do it.

  Thereafter, the browsing control unit 13 causes the page feature value forming unit 14 to execute the feature value forming process of the Web page regardless of the display (browsing) (step 102). FIG. 3 shows a specific process for forming the feature amount of the Web page as described above.

  Next, the browsing control unit 13 determines whether or not an operation for adding the URL of the target Web page to the bookmark has been performed by the user (step 103).

  The browsing control unit 13 performs such determination at, for example, the following timing. For example, it is performed at the timing when browsing (display) of the Web page is ended. The timing to end corresponds to the case where the display is switched to another Web page or the operation of closing the browser is performed. Note that the addition operation may be recognized at the timing when the addition operation is actually performed, and the addition operation may not be performed at the timing when the browsing (display) of the Web page is ended.

  When the URL is added to the bookmark, the browsing control unit 13 stores the page feature amount in the page feature amount storage unit 15 as the feature amount of the Web page that the user likes (interests), and the URL is the bookmark. If not added, the page feature value is stored in the page feature value storage unit 15 as a feature value of the Web page that the user dislikes (no interest) (step 104), and a series of processes shown in FIG. Exit.

  Next, a specific process for forming the feature amount of the Web page will be described with reference to FIG.

  In the case of the first embodiment, the feature amount of the Web page is a set of the number of appearances or the appearance rate for each token such as a word (for example, a noun) or a word string (for example, a noun phrase), and FIG. This is a process for extracting the feature amount of a simple Web page.

  When the process is started, one token is to be extracted from the head side of the Web page (HTML document) to be processed, and it is determined whether or not the token has been extracted (step 201).

  If it can be extracted, it is determined whether or not the extracted token is extracted for the first time (step 202). In other words, it is determined whether or not the currently extracted token is already registered as an element of the feature amount of the page. If the token is already registered, the number of appearances is incremented by 1 (step 203), and the process returns to step 201 described above. On the other hand, if it is extracted for the first time, it is added as a token of the feature element of the page (step 204), and the process returns to step 201 described above. When a new token is added, the number of appearances is set to 1.

  If the processing in steps 201 to 204 is repeatedly executed from the top side of the Web page (HTML document) to be processed, the token cannot be extracted before long. At this time, the creation process of the feature amount of the Web page is ended. If the frequency information for each token is not the number of appearances but the appearance rate, when the token cannot be extracted, the total number of appearances of each token is calculated, and then the processing for calculating the appearance rate is performed. Do.

  The user preference learning operation executed in parallel with the browsing of a general Web page has been described above.

  Such learning data is used when searching for Web pages. For example, it is used when determining whether or not a Web page that is a search candidate is of interest to the user. First, a feature amount is also created for such a Web page that has been determined. Thereafter, the feature quantity of the Web page that is the determination target is close to a group of feature quantities of a plurality of Web pages that the user stores (interests) that the user stores in the page feature quantity storage unit 15 as viewed statistically. Or whether it is close to a group of feature values of a plurality of Web pages that the user stores in the page feature value storage unit 15 dislikes (not interested), and determines whether the user is interested or not. . Since the method described in Non-Patent Document 2 can be applied to the statistical method for determination, detailed description thereof is omitted here.

(A-3) Effect of First Embodiment According to the first embodiment, when learning user preferences from a Web page, the user's likes and dislikes (presence of interest) for the target Web page are bookmarked. Since the determination is made based on the presence / absence of registration in the URL, the operation (work) that the user performs only to teach the system about the likes and dislikes of the Web page can be eliminated.

(B) Second Embodiment Next, a second embodiment in which the document preference learning system, method and program according to the present invention are applied to a Web page browsing system, method and program will be described in detail with reference to the drawings. .

  FIG. 4 is a block diagram showing a functional configuration of the Web page browsing system in the second embodiment. The same and corresponding parts as those in FIG. 1 according to the first embodiment are assigned the same and corresponding reference numerals. Show.

  In FIG. 4, the web page browsing system 1A of the second embodiment includes an input unit 10, a display unit 11, a communication unit 12, a browsing control unit 13A, a page feature value forming unit 14, a page feature value storage unit 15, and a preference document. A determination unit 16 and a browsing time counting unit 18 are included. Although the bookmark storage unit 17 exists, it is not shown in FIG. 4 because it is away from the features of the second embodiment.

  The browsing time counting unit 18 counts the time during which a certain web page is being browsed by the user (the stay time at the URL) under the control of the browsing control unit 13A. The browsing time counting unit 18 counts, for example, the time from the start point at which a certain Web page is displayed on the display unit 11 to the point at which the display is ended as the browsing time. The end time corresponds to a time point when switching to display of another Web page or a time point when an operation of closing the browser is performed.

  In this second embodiment, the browsing time (stay time at URL) of the Web page is automatically set to the user's likes and dislikes (presence / absence of interest) based on the fact that the user's high interest page is longer than the low page. It is adopted as an index to grasp.

  FIG. 5 shows the flow of the entire operation when learning the user's document preference for a certain Web page regarding the second embodiment, and the same steps as FIG. 2 according to the first embodiment include These are shown with the same reference numerals.

  In the case of the second embodiment, as shown in FIG. 5, the step 303 for determining whether the user likes or dislikes is performed based on whether or not the browsing time exceeds the threshold, not the presence / absence of bookmark registration (step 103). It has become. Other processing steps are the same as those in the first embodiment.

  When the browsing time exceeds the threshold, the browsing control unit 13A of the second embodiment likes the feature amount of the target Web page in the page feature amount storage unit 15 (interested). If the browsing time is less than or equal to the threshold value, the feature amount of the target web page is stored in the page feature amount storage unit 15 as the feature amount of the web page. As a quantity (step 104), the series of processing shown in FIG.

  According to the second embodiment, when learning user preferences from a web page, the user's likes and dislikes (presence of interest) for the target web page are determined based on the length of stay in the URL. Therefore, it is possible to eliminate the operation (work) that the user performs only to teach the system about the likes and dislikes of the Web page.

(C) Other Embodiments The method for automatically determining the user's preference in each of the above embodiments is not only applied to the web page browsing system as a single unit as in the embodiment, but also in combination with other methods. You may make it apply to a page browsing system.

  For example, the automatic determination mode can be set, and when the automatic determination mode is on, the determination method of each of the above embodiments is applied. When the automatic determination mode is off, the “like” icon or A “dislike” icon may be displayed and designated by the user.

  In addition, for example, a method of displaying a “like” icon or a “dislike” icon is applied until the number of pages in which the feature amount is stored in the page feature amount storage unit 15 becomes a predetermined number of pages. The determination methods of the above embodiments may be applied.

  Further, for example, the methods of the first and second embodiments may be combined. For example, it may be determined that a Web page whose URL is bookmarked is immediately liked, and for those that are not bookmarked, the browsing time may be confirmed to determine whether or not the user likes or dislikes.

  In each of the above-described embodiments, what is determined by two categories of likes and dislikes is shown, but the number of categories may be increased. For example, when bookmarks can be registered in any of a plurality of folders, a plurality of sub-classes may be provided for likes. Further, for example, the browsing time may be divided into three or more.

  In each of the above embodiments, in the learning mode, all types of Web pages are determined as likes and dislikes, but depending on the types, they may be excluded from likes and dislikes. For example, Web pages in which the area of an image (including moving images) exceeds a predetermined percentage of the total area may be excluded, and Web pages in which the total number of tokens in the Web page is equal to or less than a predetermined number are excluded. You may make it.

  In the first embodiment, the Web page whose URL is registered in the bookmark is determined to be liked. However, for later processing, the user saves the Web page being browsed as a file (file In the case of registration), the user may treat the Web page as likes. That is, it is determined that the user likes the Web page in which the user has performed the act of saving at least part of the information related to the Web page for future processing and use.

In the second embodiment, the threshold for determining likes and dislikes according to the viewing time is a fixed value. However, the user may be able to variably set the threshold. Further, the system may select a threshold value from a plurality of types according to the type of the Web page. For example, the threshold value may be automatically switched between a top page of a company or the like and a lower Web page in a tree structure hierarchy from the top page. Whether the page is a top page can be determined by the presence of a predetermined keyword, the structure of a URL, and the like.

  As for the second embodiment, when a user moves to another web page after browsing a certain web page and then returns to the first web page (the one stored in the cache). It may be handled as follows. If it is determined that the user likes the first browsing, the subsequent browsing is excluded from the determination. If it is determined that the user dislikes the first browsing, (1) the browsing time after returning is also included in the browsing time and re-determination is performed, or (2) the viewing time is displayed again regardless of the initial browsing. Re-determine based on. In both cases (1) and (2), when a determination is made that the re-display is reflected, the initial dislike information is deleted from the page feature amount storage unit.

  The method described in each of the above embodiments uses a line such as a personal computer connected to the Internet, a mobile terminal device (for example, a mobile phone, a PDA) or the like, a LAN, a telephone line, a dedicated line, or a wireless network. It can be used from the constructed communication network, and can provide information collection for optimum preference determination to any system.

  In each of the above embodiments, the Web has been described as an example. However, the present invention is not limited to the Web, and any communication application can be used. For example, electronic mail can be used.

  In the present invention, any document can be targeted. Therefore, it is not always a requirement to perform communication via the network. For example, it is possible to target a document read from a storage medium such as a CD-ROM.

It is a block diagram which shows the functional structure of the web page browsing system in 1st Embodiment. It is a flowchart which shows the flow of the whole operation | movement in 1st Embodiment at the time of learning a user's document preference about a certain web page. It is a flowchart which shows the flow of the production | generation operation | movement of the page feature-value in 1st Embodiment. It is a block diagram which shows the functional structure of the web page browsing system in 2nd Embodiment. It is a flowchart which shows the flow of the whole operation | movement in 2nd Embodiment at the time of learning a user's document preference about a certain web page.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1, 1A ... Web page browsing system, 10 ... Input part, 11 ... Display part, 12 ... Communication part, 13, 13A ... Browsing control part, 14 ... Page feature-value formation part, 15 ... Page feature-value memory | storage part, 16 ... Preference document determination unit, 17 ... bookmark storage unit, 18 ... browsing time counter, 20 ... Internet.

Claims (6)

A feature quantity acquisition means for obtaining a feature quantity of the input document;
It is determined whether or not an action to save the information related to the input document for the future use and use of the input document has been performed, and it is determined that the action to be stored is a document in which the user is interested 1 As a condition, interest discriminating means for discriminating whether the user is interested in the input document,
A document preference learning system comprising: a document feature amount storage unit that stores a feature amount of the input document according to whether the document is a document of interest to a user. A feature quantity acquisition means for obtaining a feature quantity of the input document;
An interest discriminating means for discriminating whether or not the user is interested in the input document, as one condition for discriminating that the user's browsing time of the input document exceeds a threshold as a document of interest to the user;
A document preference learning system comprising: a document feature amount storage unit that stores a feature amount of the input document according to whether the document is a document of interest to a user. A feature amount acquisition unit, an interest determination unit, and a document feature amount storage unit;
The feature amount acquisition means obtains the feature amount of the input document,
The interest discriminating unit determines whether or not an act of storing the information related to the input document for future use or use of the input document has been performed, and the user is interested in performing the storing operation. As one condition for discriminating from a certain document, the user's interest in the input document is discriminated,
The document feature amount storage means stores the feature amount of the input document according to whether the document is a document of interest to the user or not. A feature amount acquisition unit, an interest determination unit, and a document feature amount storage unit;
The feature amount acquisition means obtains the feature amount of the input document,
The interest determination means determines whether the user is interested in the input document as one condition for determining that the time when the user is browsing the input document exceeds a threshold as a document of interest to the user,
The document feature amount storage means stores the feature amount of the input document according to whether the document is a document of interest to the user or not. Computer
A feature quantity acquisition means for obtaining a feature quantity of the input document;
It is determined whether or not an action to save the information related to the input document for the future use and use of the input document has been performed, and it is determined that the action to be stored is a document in which the user is interested 1 As a condition, interest discriminating means for discriminating whether the user is interested in the input document,
A document preference learning program characterized in that it is described so as to function as a document feature amount storage unit that stores the feature amount of the input document according to whether the document is a document of interest to the user. Computer
A feature quantity acquisition means for obtaining a feature quantity of the input document;
An interest discriminating means for discriminating whether or not the user is interested in the input document, as one condition for discriminating that the user's browsing time of the input document exceeds a threshold as a document of interest to the user;
A document preference learning program characterized in that it is described so as to function as a document feature amount storage unit that stores the feature amount of the input document according to whether the document is a document of interest to the user.

Images