JP2002197115A - Web page retrieval method with usage of evaluation reference data and storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To properly retrieve a Web page by setting minute conditions for each topic of retrieval, easily set the minute conditions required for the retrieval, and adaptively improve retrieval efficiency when retrieval result is insufficient. SOLUTION: With the usage of evaluation reference data wherein minute conditions can be set for each of the topic of the retrieval, interest of the user can be delicately and properly reflected in a retrieval process. The evaluation reference data is stored in a storage device, thereby making the evaluation reference data easily reusable. By evaluating a link of the Web page with the evaluation reference data, the Web page desirable as a retrieval object is given priority to be read. Thus, effective and proper retrieval can be achieved. When the retrieval result is insufficient, a common attribute of the insufficient Web page is picked out, and a condition not to fulfill the common attribute is newly added. Thus, the evaluation reference data can adaptively updated.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for retrieving a web page (that is, a document readable on the Internet) and a storage medium storing a program according to the method. By using the evaluation criteria data describing the conditions and making this evaluation criteria data reusable, it is possible to perform a search accurately and easily, for example, the same level as a conventional search engine Despite its simplicity, it is necessary to perform a more accurate search than a conventional search engine, to efficiently search even web pages that are not registered in the search engine database, or It replaces existing search procedures with inexpensive automatic means, and is useful for sharing information, which is a feature of the Internet. It is intended to further promote the passage.

[0002]

2. Description of the Related Art A typical conventional technique for collecting information of a web page is a method using a so-called search engine web robot. This is done by running a number of programs that automatically load web pages,
This is a system that collects information of many web pages at high speed and comprehensively and creates a database. However, this method is not suitable for general users to collect information focused on specific interests, because it is a large-scale and comprehensive information collection method and is expensive.

[0003] In order for ordinary users to easily search web pages, the most widely used method at present is to use a "search engine" which is a service on the Internet. For example, Internet Expl
Yahoo! using a web browser such as orer (trademark of Microsoft Corporation). By using a search function provided by a search engine such as that described above, a web page can be searched easily and quickly. However, as is well known,
There is a problem that a search result by this method includes many web pages unrelated to a search target intended by a general user. Also, in this method, since there are great restrictions on the search conditions that can be specified, in many cases, it is often impossible to specify the detailed conditions necessary to avoid the irrelevant web page as described above. For example, if an individual expresses their interests related to their hobbies in detail, the number of sheets may be as many as 400 characters in a manuscript sheet, but a current search engine specifies a condition of more than one hundred characters. That is not practical. Since there is already a limit at the time of inputting the search condition, it is inevitable that the search result of the conventional search engine will be somewhat inaccurate.

As a more effective search method of a web page than a simple search engine search, a method of using a “meta search” for simultaneously requesting a plurality of search engines is well known. This is a search software, which has the advantage that it has a function of inquiring many search engines at the same time, so that there is a greater chance of finding web pages that meet the search conditions. However, this method is basically a method of merely collecting search results of search engines, and the search results reflect the limitations of the search engines themselves. In other words, the search results inevitably include a lot of information irrelevant to the user's interest, and are not suitable for a high-quality search that reflects even a small point of the user's interest. Also, as with a single search engine, users have to spend a lot of time reading their own web pages to find the information they are looking for from the incorrect search results. Therefore, it is troublesome.

There is also a method of preparing a specialized database in order to quickly perform an accurate search. This has been put to practical use, for example, as a search system for medical literature and a search system for patent information. If it is limited to a specific field, by constructing such an exhaustive database and providing a search means suitable for a topic, it becomes possible to search accurately and quickly. If this method can be applied to web pages as it is, the above problems will be solved, but the size of the Internet is orders of magnitude larger than information sources such as medical literature, and it is updated daily, making it difficult to investigate and track Therefore, it is very difficult to make a database of the entire web page. For example, even search engines, which are said to cover the widest range at present, can at most only tens of percent of the Internet as a database. Further, in the case of the Internet, it is necessary to deal with information in all fields, so it is difficult to create a detailed index for each specialized field, and thus it is also difficult to perform an accurate search using the detailed index. For example, in the index of the medical literature database, detailed items such as the publication date of the literature, the author, the publisher, and the like can be set, so that accurate search can be performed using these conditions. However, web pages are so non-limiting that it is difficult to use such special conditions as database indexes. As described above, in the case of web pages, it is difficult to achieve completeness and detailed index, which are essential conditions for specialized databases, so there are problems in applying specialized database search methods. Many.

As described above, it is difficult to accurately search a general web page by a mechanical method using a search engine or the like, and a manual search service is provided as a business. This is a service in which a search operation performed by a user is usually performed by an expert, and an accurate search is possible because a human reads and checks a web page. However, since this search method depends on humans,
It requires a great deal of cost and time, and is not suitable for general users to easily search web pages.

As described above, regardless of whether the method is a mechanical method or a method requiring manpower, the conventional web page search method has the first condition that general users can easily use the method, Nothing simultaneously satisfies the second condition of performing an accurate search that reflects the interests of the above. For an accurate search, detailed conditions must be considered in order to reflect the user's interest in the topic to be searched. However, a mechanical search method or search for specifying such detailed conditions must be considered. The lack of equipment is the essential point of this problem. Even if a mechanical search method or a search device capable of specifying detailed conditions as described above was created in order to perform an accurate search,
At the same time, unless a simple method or device for specifying the various conditions is also provided at the same time, the user needs to input a large number of conditions for each search, which is very troublesome and impractical.

[0008]

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned various problems or problems, and enables detailed setting of conditions for each search topic.
An object of the present invention is to provide a method of searching for a web page accurately without requiring a person to perform a search, and a storage medium storing a program according to the method, by making the condition setting easily reusable.

[0009]

According to the present invention, there is provided a method for retrieving a web page, which is used as a reference for evaluating a web page when performing a search on a specific topic in response to a user request. A first step of creating evaluation criteria data by the user or a third party; and a second step of designating evaluation criteria data corresponding to a search topic from reusable evaluation criteria data already created.
A third step of determining the evaluation reference data at the time of execution by the user inputting search conditions; and a third step of setting a set of web pages waiting to be read at the time of starting to read web pages. 4 and steps, from the set of the read waiting web page, a fifth step of selecting the highest web page evaluation calculated based on the evaluation reference data during the execution as the next read target, the fifth A sixth step of reading the web page selected in the step of; and a web page referenced by a link in the read web page,
A seventh step of adding a candidate to be read in the future to the set of web pages waiting to be read, and an eighth step of calculating an evaluation of the read page based on the evaluation reference data at the time of execution. I do. As an example, after the eighth step, the method includes a step of displaying the read web pages to the user as a list (hereinafter, referred to as a search result list) in descending order of evaluation. Further, as an example, the method includes a display processing step in which a user can instantly browse a corresponding web page by designating one item of the search result list with a mouse click or the like. In order to cope with a case where the search result list is not satisfactory to the user, the evaluation reference data may be adaptively updated using the search result list. As an example, it is possible to designate a satisfactory page and an unsatisfied page of the user from the search result list, set a common attribute of the satisfied page as an attribute 1, set a common attribute of the unsatisfied page as an attribute 2,
A step of adding both the condition of “satisfying the attribute 1” and the condition of “not satisfying the attribute 2” to the evaluation criterion data.

The outline of the method according to the present invention will be described. First, before a searcher equipped with the search method of the present invention starts a search, evaluation criterion data serving as a criterion for ranking web pages in a search process is created and stored in a storage device (first example). Steps). This evaluation criterion data is a condition that becomes a criterion when evaluating a web page, and includes a determination algorithm for determining whether the web page satisfies the condition, or a condition that the web page satisfies the condition. Any algorithm may be used as long as it has an algorithm for calculating the degree of the numerical value or vector. The determination result of the web page based on the evaluation criterion data, or a numerical value or a vector indicating the degree to which the web page satisfies the evaluation criterion data, is hereinafter referred to as “web page evaluation” for convenience.
The data format of the evaluation criterion data may be any format that can be used by the searcher. In the search process, the evaluation criterion data functions as data describing the user's interest in the search target. That is, a web page with a higher evaluation in the light of the evaluation reference data is estimated to be data of higher interest to the user, and is ranked higher in the search results. Further, the evaluation reference data can be created by a method of inputting data by a user of the searcher, inputting data by a third party, or automatically generating the data by software. Further, the evaluation criterion data may be created at any time prior to the start of the search. By appropriately storing the evaluation criterion data created for each search topic in a storage device, the evaluation criterion data can be read out and easily reused when searching for the same topic. This evaluation criterion data may, of course, be used by an individual or shared with a third party on a network.

[0011] Before the searcher begins searching, specifying one of the evaluation criteria data (second step). The specification of the evaluation criterion data may be performed by the user of the search device by inputting to an input screen prepared by the search device,
The search may be performed by automatically selecting according to a predetermined algorithm from evaluation reference data in a storage device or evaluation reference data that can be read on a communication network. For example, after a third step in which a search condition is specified as described later, evaluation reference data of a topic most closely related to the search condition may be automatically selected programmatically. Specifically, each evaluation criterion data is saved under a file name representing each topic, and when a search condition composed of a single keyword is the same as the topic name, the evaluation name of the topic name is stored. For example, there is a method of automatically selecting evaluation reference data.

A user of the searcher inputs search conditions (third step). The search condition is a condition that specifies the search target itself while the evaluation criterion data determines a condition that is the background of the search. However, when it is considered that the user has also specified the search target by specifying the evaluation reference data, the input of the search condition may be omitted. The search condition can be specified by a logical combination of keywords similar to that of a conventional search engine, or any other expression form that can be interpreted by a searcher. For example, if the searcher has a function of natural language analysis, it is possible to specify search conditions in natural language. When the search condition is input, evaluation reference data actually used in the search process (hereinafter referred to as “execution evaluation reference data”) is generated from the search condition and the evaluation reference data. The generation of the evaluation reference data at the time of execution refers to an operation of putting together the search condition and the evaluation reference data, which are two data serving as a reference for evaluating a web page. Although a detailed description of this process will be described later, for example, the search condition may be regarded as one indispensable condition and added to the evaluation criterion data to be put together. If the user of the searcher omits the input of the search condition, the evaluation criterion data is determined to define the search condition at the same time. Therefore, the evaluation criterion data is not changed. Instead, it may be used as it is as the evaluation reference data at the time of execution.

In order to start reading a web page, a set of web pages waiting to be read is initialized (fourth step). Here, hereinafter, a set of web pages waiting to be read will be referred to as a “ready set” for convenience. U of web page to be loaded in the future
As long as the set of RLs can be clearly specified, the set waiting to be read may take any data format. Here, the URL is, as is well known, address information that uniquely specifies a web page on the Internet. The initialization of the set waiting for reading may be appropriately performed by an arbitrary method. For example, using the search condition specified by the user in the third step, Yahoo! Or a search request to Google or the like, and the top 50 URLs of the search result are set as a read waiting set.

The web page with the highest evaluation calculated based on the evaluation reference data at the time of execution is selected as the next read target from the read waiting set (fifth step). The calculation of the rating here has to be based on information available about the web page before actually reading the web page. For example, in the case of a web page included in the set waiting to be read which is initially set by using the search engine in the fourth step, the ranking or score in the search result can be used as it is. A more general case is a case where the web page is added to the set waiting to be read in a seventh step described later, which will be described later in a section describing the seventh step.

The web page selected in the fifth step is read (sixth step). When the reading of the web page is started, the web page is removed from the set waiting for reading. In order to improve the efficiency of the search, the fifth step and the sixth step may be repeated one after another while reading the web page, so that a plurality of web pages are read simultaneously. This can be realized by using a multi-thread function of a computer or a plurality of machines. Here, the multi-thread function of a computer is a function of the computer that enables multiple processes to be performed in parallel in one program, and when one process is delayed, it is calculated in another process. Since the resource can be controlled to be used, an efficient program can be created. For example, using a programming language such as Java, a program using a multi-thread function can be created relatively easily.

The web page read in the sixth step generally contains a plurality of links.
Among the web pages referred to by those links, a web page that is a candidate to be read in the future is newly added to the set waiting to be read (seventh step). Here, the link in the web page is, as is well known, a reference to another web page described in the web page.
When a link displayed on a browser such as Internet Explorer (trademark of Microsoft Corporation) is clicked with a mouse, a web page referred to by the link is automatically read and displayed on the browser. What is used. Whether one web page is a candidate to be read may be determined in various ways. For example, the web page may be simply a candidate to be read on condition that the web page has not been read yet, or after obtaining a clue about the content of the web page in some way, the web page is promising as a search target. It may be determined whether or not there is, and only when it is promising, may be added to the read waiting set. As an example, a link on one web page is highly evaluated by the evaluation criterion data at the time of execution by utilizing the property that a link on a web page is often referred to a web page related to the content of the web page. There is also a method of presuming that the web page referred to by the link of the web page is a promising search target. With respect to the web page added to the set waiting to be read in the seventh step, the web page (the loaded web page) on which the link referring to the web page has been posted is used in the fifth step. Perform an evaluation. That is, the web page (the web page waiting to be read)
Is analyzed from the description of the web page (the loaded web page) having the link, based on the evaluation criteria data at the time of execution. And calculate the evaluation, and use the calculated value as the evaluation of the web page. Specific examples of the analysis and evaluation will be described later in the section of the embodiment of the invention.

The evaluation of the web page read in the sixth step is calculated based on the evaluation reference data at the time of execution (eighth step). For the convenience of the user of the searcher, a set of loaded web pages may be displayed in a list in descending order of evaluation.
Further, it is convenient to be able to easily refer to attribute information such as titles, summaries, and evaluations of individual web pages displayed in a list.

As described above, according to the present invention, attention is paid to the problem of condition setting in search, and evaluation reference data capable of finely handling each search topic is considered and stored in a storage device. Is easily reusable, and based on this evaluation criterion data, the potential of a web page as a search target is determined, and a promising web page is read first and searched. You can easily set detailed conditions that were impossible, search with high efficiency by avoiding irrelevant web pages, and do not need a dedicated database,
There is no need to rely on manual labor that requires a great deal of cost and time, and therefore, it has excellent effects such as being optimal for a general user to easily and accurately search a web page.

By the way, since the user of the searcher has made a search request by creating evaluation reference data that does not accurately reflect his / her interest, the search result may be unsatisfactory to the user. To address this problem, the content of the evaluation criterion data may be adaptively updated using the search result so as to accurately reflect the user's interest. That is, the method further includes a step of specifying an undesired attribute of the web page which the user considers unsatisfactory by using a predetermined algorithm, and then adding a condition of “not having the attribute” to the evaluation criterion data. It shall be. In addition, as described later, by extracting “desired attributes” from a web page satisfied by the user, the evaluation criterion data can be updated so as to more accurately reflect the interests of the user. As described above, when the evaluation reference data is automatically and adaptively updated, it is preferable to further include a step of confirming the update with the user before the update in order to eliminate the possibility of an inappropriate update.

When the user is interested in a plurality of topics and has evaluation criteria data for each topic, the search method of the present invention is used to bookmark the browser (Internet Exp. Of Microsoft Corporation).
It is also possible to perform automatic updating of the "favorite" in the lorer. For example, an automatic search is periodically performed for all topics, and the top 30 search results are obtained.
It is sufficient to update the bookmark so that the matter is a component of the topic's folder in the bookmark. Specifically, since the bookmark is stored in the computer as a file in a special format, an update process for this file may be performed. Such an application to the automatic update function of the bookmark is an application example depending on the accuracy of the search result, and when there are many web pages irrelevant to the user's interest like the search result by the conventional search engine, However, the search result is not suitable for use as a bookmark as it is.

The present invention can be implemented not only as a method invention but also in the form of a storage medium storing a program according to the method. That is, the computer-readable storage medium according to the present invention stores a program for causing the computer to execute a web page search method based on evaluation criterion data, and the program includes a program for evaluating a web page. A first step in which a user or a third party creates reusable evaluation criterion data as a reference, and an evaluation criterion corresponding to a search topic from among the reusable evaluation criterion data already created. A second step of designating the data, a third step of determining the evaluation reference data at the time of execution by the user of the storage medium by inputting search conditions, and reading at the time of starting reading the web page From the fourth step of setting a set of waiting web pages and a set of waiting web pages, A fifth step of selecting a web page with the highest rating calculated based on the evaluation criterion data as a next read target, a sixth step of reading the selected web page, and a link in the read web page. A seventh step of adding a web page that is a candidate to be read in the future to the read web page to a set of web pages waiting to be read, and calculating an evaluation of the loaded page based on the evaluation reference data at the time of execution. 8 steps.

[0022]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a functional block diagram illustrating the principle of a method for searching a web page according to the present invention. Block 10 indicates an information source readable on the Internet, or a computer storing the information source. Denotes a searcher, that is, a device provided with the search method of the present invention. In the block of the searcher 20, the processing executed for searching the web page is conceptually shown by functional blocks. As will be described later, some of these functional blocks may be appropriately omitted. Further, in the example described below, each functional block is configured separately, but may be configured integrally.

FIG. 2 is a flowchart showing an example of a search program according to the search method according to the present invention, and the following description will be made with reference to this flowchart. First, prior to a search for a web page, evaluation reference data for ranking web pages is created (step ST1). Here, a web page is a document that can be read from the Internet and displayed by a browser, as is well known.
In addition to text, multimedia files such as images can be posted on web pages. Here, the ranking of the web pages means to determine the order of priority when reading the web pages or to determine the order when displaying the loaded web pages in a list as described later.

FIG. 3 is a schematic diagram showing an example of a method for evaluating a web page and a link based on the evaluation reference data.
As described above, the evaluation criterion data is data describing a condition that is a criterion for evaluating a web page, and includes a determination algorithm for determining whether or not a specific web page satisfies the condition, or Any algorithm may be used as long as there is an algorithm for calculating how much the page satisfies the condition using a numerical value or a vector. For example, a condition that “the web page has a file size of less than 10 kilobytes” or a condition that “the web page includes five or more images” can be considered. The evaluation reference data at the upper left of FIG. 3 is an example of the evaluation reference data, and is composed of a list of keywords. For each keyword of the evaluation criterion data, it is possible to set the necessity and weight. Here, the necessity is a condition for determining whether or not the web page needs to include the keyword, and two types of necessity can be specified. Here, the weight is a numerical value that is an index of the importance of the keyword, and is a condition that determines how much the inclusion of the keyword in the web page contributes to the evaluation of the web page. is there. A method of calculating the evaluation using the weight will be described later. An example of the evaluation criterion data is stored in a computer as evaluation criterion data on the topic “music”, so that it can be reused when a search on the topic “music” is performed again. Also, in this example, there are only three keywords. However, as described above, the search method of the present invention shows its true value using complex evaluation standard data in which the number of keywords ranges from tens to hundreds or more. This is the case.

The upper right part of FIG. 3 conceptually shows an example of a web page evaluation method when an example of the evaluation reference data is used. If the web page does not satisfy the required condition (that is, does not include the required keyword “music”), the rating = 0.
Set to 0. For web pages that meet the requirements,
The number of appearances S of the keyword on the web page is counted for each keyword, and a predetermined monotonically increasing function f
Is used to calculate the function value f (S), and the product of the keyword with the weight W, that is, W × f (S) is calculated. Then, a numerical value {W × f (S)} obtained by summing the calculated values for all the keywords is used as the evaluation of the web page. In other words, the calculation method has a property that the evaluation is higher as the function f includes more keywords due to monotonic increase, and the evaluation is higher as the function f includes a keyword having a larger weight because it takes a product with W. .

One of the limitations of the conventional search engine is that it is not possible to specify a search condition with a keyword weight as described above. For example, if a user is interested in protecting the copyright of music, and this user searches for the keyword "music" and the keyword "copyright", a web page for music fans who only touches the copyright very little Even so, the search condition is well matched. However, using the above-described evaluation reference data as an example of the present invention, the keyword “music” is indispensable and has a weight of 1.
If 0 and the keyword “copyright” are required and the weight is specified as 9.0, the page that places more importance on the copyright is highly evaluated, so that a more accurate search result can be expected.

The creation of the evaluation reference data in step ST1 of FIG. 3 may be performed by an appropriate method. For example, it is also conceivable to use a distributed text editor (program for editing a document file) or a binary editor (program for editing a binary file) to directly edit a file describing evaluation reference data. But,
When the evaluation criterion data has a complicated format, it is troublesome to create and edit the evaluation criterion data with the distribution editor. Therefore, it is preferable that the searcher itself has a function for easily editing the evaluation criterion data. The block of the evaluation criterion data editor 21 in FIG. 1 conceptually shows the creation / edit processing.

Next, the evaluation reference data used in the search is specified (step ST2). The specified evaluation criterion data may be data stored inside the searcher (illustrated by an ellipse located immediately below block 21 in FIG. 1), or data readable from the Internet (block 10 in FIG. 1). (Illustrated by an ellipse located on the right side of the inside). Also, the creator of the evaluation reference data may be anyone. Evaluation criteria data selector 22 in FIG.
Block conceptually shows this designation processing. This designation process may be performed by the user of the searcher inputting from an input screen provided by the program, or may be performed by another arbitrary method. In short, it is sufficient to determine one evaluation criterion data to be used in the search before starting the search.

Next, search conditions are input and evaluation reference data at the time of execution are generated (step ST3). The block of the search condition input device 23 in FIG. 1 conceptually shows the input process of the search condition. The elliptical figure of “evaluation reference data at the time of execution” indicated by the arrow coming out of the block 23 is: This conceptually shows the evaluation reference data generated in step ST3. As described above, if the user of the searcher considers that the evaluation reference data specified in step ST2 also serves as a search condition, step ST3 can be omitted, and the evaluation reference data at the time of execution can be omitted. The evaluation reference data itself specified in step ST2 may be used. The input format of the search condition may be any format that can be interpreted by the search condition input device. Here, being interpretable means that the searcher has an algorithm for determining whether or not the web page satisfies the search condition. For example, the fact that the search condition input device is capable of interpreting the logical combination of keywords means that a search condition of “Mozart AND Requiem” can be input, and the search device provides a “Mozart AND Requiem” "And" Requiem "have the function of determining whether or not both keywords are included.

The step ST3 need not always be performed after the step ST2, but may be performed simultaneously with the step ST2 or before the step ST2. For example, after the search condition input by the user of the searcher (an example of step ST3), the searcher automatically specifies the evaluation reference data of the topic most closely related to the search condition (an example of step ST2). May be followed. The evaluation criterion data and the search conditions do not have a clear distinction in content in that both serve as the criterion for the evaluation of the web page, but rather are distinguished by the operational role of the search processing procedure. That is, as described above, the evaluation criterion data specifies the background of the search, and is usually reusable data composed of a large number of conditions, while the search condition specifies the search target and is usually a small Is disposable data composed of the following conditions. When evaluating the web page in search process, based on these two types of data, one of the evaluation is made. In other words, when the search is executed, these two types of data function as one evaluation criterion as a whole. Since this one evaluation criterion must be able to calculate the evaluation of the web page based on it, it is, by definition, itself one evaluation criterion data. This is nothing but the evaluation reference data at the time of execution described above. However, it is not always necessary to create a new entity integrating the search conditions and the evaluation reference data in the storage device. It is also possible to always manage the search conditions and the evaluation reference data separately inside the searcher. is there. However, it is considered that in many cases, it is appropriate to create and store the evaluation reference data at the time of execution as temporary data until the end of the search separately from the search conditions and the evaluation reference data. Can be This is because the two evaluation criteria of the search condition and the evaluation reference data are combined into one to make it easier to handle programmatically, as well as the characteristics of the two types of storage devices of the computer, the hard disk and main memory. That's why. That is, the evaluation reference data that requires a storage capacity for long-term storage, reuse, and storage of a large number may be stored as a file in a hard disk that has a large capacity and is suitable for long-term storage. On the other hand, just use temporarily,
The evaluation reference data at the time of execution that does not reuse and requires a small amount of storage capacity but requires high-speed reading of data is stored as data in the main memory that is small in size but can be accessed at high speed. Good. From such a viewpoint, it is advisable to take an embodiment in which the evaluation reference data in the hard disk and the input search condition are added to generate the evaluation reference data at the time of execution in the main memory. It can be said that. Further, the evaluation reference data in the hard disk and the evaluation reference data at the time of execution do not necessarily have to be evaluation reference data in the same format.

After the evaluation reference data at the time of execution is determined, a set waiting to be read is initialized (step ST).
4). As described above, the initial setting is, for example, Ya
hoo! And a search engine such as Google. For example, when the searcher user specifies a search condition of “Mozart AND Requiem”, first, Yahoo! Then, a search request is made under the search condition, and the top 20 web pages are extracted from the search result, and the web page may be set as a read waiting set. This example can easily be implemented purely programmatically (without user intervention).

Next, the web page with the highest evaluation is selected from the set waiting for reading (step ST5). The block of the read target selector 24 in FIG. 1 conceptually illustrates this selection processing. This step ST5
Corresponds to step ST, as can be seen from the flowchart of FIG.
6, ST7 and ST9 form a loop. The processing in step ST5 immediately after step ST4 is slightly different from the processing in step ST5 after circulating through the loop, as described below. Immediately after step ST4, for web pages belonging to the set waiting to be read,
Limited information is available for evaluation. For example,
As in the previous example, Yahoo! Such information is provided by a search engine database. Therefore, the method of evaluating a web page is naturally limited. For example, Y
ahoo! The highest rated web page is selected by a method such as using the rank of the search result as the ranking of the web page as it is.

On the other hand, in the seventh step, when a web page to which a link in the actually read web page is referred to is added to the set waiting to be read, the read web page is used. And evaluate. This evaluation method is performed by analyzing the description on the read web page as described above. As an example of this analysis method, there is a method of using sentences before and after a link. That is, since sentences before and after a link that refers to the web page are generally considered to be closely related to the web page, the link and the sentences before and after the link (that is, a portion near the link) are substituted for the web page. Web page. Then, the evaluation of the virtual web page is calculated based on the evaluation criterion data at the time of execution, and is set as the evaluation of the web page. FIG.
Is a schematic diagram conceptually showing this evaluation method. The left side of is a conceptual diagram showing a state in which a link is posted in a web page, and the right side is a conceptual diagram showing a neighborhood extracted from sentences before and after the link. here,
As for the text before and after the link, it would be effective to limit the range so as not to include another link existing before and after the link on the web page where the link is described. This is because, for example, a link in a web page in which many links such as a menu on a web page screen and a link collection are arranged in a row tends to be unrelated to the text before and after the link. As another example (improvement example) of the analysis method, there is a method of using a sentence structure before and after a link. That is, if the link itself is represented as a character string, the character string is represented as an image, and if the link is represented as an image, the description of the image (technically, H
An evaluation algorithm that places the highest priority on the character string described in the alt attribute in the img tag of the TML file. In addition, the evaluation algorithm is such that the sentence before the link is more important than the sentence after the link. As described above, it is preferable to devise an algorithm for more accurate evaluation by utilizing the structural characteristics of the web page.

Next, the web page selected from the set waiting for reading is read (step ST6). The read web page is temporarily stored in a storage device. The block of the reading unit 25 in FIG. 1 conceptually illustrates this reading processing. Reading a web page can be easily performed programmatically. As described above, the reading efficiency may be improved by using the “multi-thread function”. When the "multi-thread function" is used, for example, the step ST5 may be performed a plurality of times without waiting for an input from another process. Alternatively, as another example, a method is conceivable in which a plurality of web pages with the highest evaluation are selected from a set waiting to be read, and reading of the web pages is started simultaneously.

When the reading of one web page is completed, next, among the web pages referenced by the links in the web page, those which are candidates for reading in the future are added to the waiting set for reading (step ST7). The block of the link adder 26 in FIG. 1 conceptually illustrates this addition processing. Because web pages are written according to certain rules and the format for writing links is also promised,
By analyzing the loaded web page, the link information therein can be easily extracted programmatically. If the link information extracted in this way is suitable as a future read candidate, it is subjected to data format conversion or the like as appropriate, and added to a read waiting set stored inside the storage device of the searcher. I just need. Although the performance is degraded, the determination as to whether or not one web page is appropriate as a reading candidate may be omitted. However, in practice, in order to avoid wasting reading pages once and over again, it is necessary to make sure that the link does not refer to a web page that has already been read before adding it to the waiting list. Good to do.

When the reading of one web page is completed, the evaluation of the web page is calculated (step ST).
8). The block of the page evaluator 27 in FIG. 1 conceptually illustrates this evaluation calculation process. This step S
T8 does not need to be performed after step ST7, and may be performed at an appropriate time after the end of reading the web page. Further, the embodiment of the evaluation calculation of the web page has already been specifically described with reference to FIG. As described above, the read web pages may be displayed in a list in the order of their evaluation. The block of the search result list creation / display unit 28 in FIG.
7 is a diagram conceptually showing a display process. It is convenient to make the program such that, when the item of the displayed list is clicked with a mouse, the web page corresponding to the item can be immediately viewed.
In addition, for the convenience of the user, the search result list may be saved so that the user can refer to it again later.

By the way, since the number of web pages on the Internet is enormous (it is said to be billions at the time of filing this patent application), reading of web pages by a loop formed by the above steps ST5, ST6 and ST7 is performed. The process can be considered a virtually endless process.
In such a huge number of web pages, the search process of the present invention may be continued virtually endlessly. However, in many cases, it is often desirable to provide a search termination condition in advance and terminate the web page reading process and the search process when the condition is satisfied. In such a case, it is preferable to include an end condition determination processing step as in step ST9 of FIG. While the end condition is not satisfied, the above-described loop is repeated. When the end condition is satisfied, for example, a final search result list is displayed, and a search process end process is performed. The termination conditions include, for example, ending when a web page with a certain rating or more is found, ending after searching for a certain time, ending after reading a certain number of web pages, and changing the top 50 items in the search result list for a certain time. Conditions such as exiting if not done will be valid. It is desirable to flexibly set such an end condition according to the purpose of the search. For this purpose, a function that allows the user to easily set the end condition may be added to the search device.

By the way, even if the user creates the evaluation reference data by himself, the user does not always become aware of his or her own interests, so that the evaluation reference data does not accurately reflect the user's interest. It is possible. In order to cope with such inaccurate search resulting from inaccurate evaluation criterion data, the content of the evaluation criterion data may be updated adaptively (by feedback) when the search result is unsatisfactory. The block of the evaluation criterion data adaptor 29 in FIG. 1 conceptually shows this adaptation update processing. The functional block diagram of FIG.
5 is a schematic diagram illustrating an example of the adaptation update process. First,
From the search result list, a web page that the user is dissatisfied with is checked. Next, the searcher extracts conditions common to the checked pages (that is, undesired conditions) by a predetermined algorithm. Then, finally, a condition that the web page does not satisfy the condition is added to the evaluation criterion data. By doing so, in subsequent searches using this evaluation criterion data,
Inappropriate pages having the common attribute are automatically excluded. Conversely, even when the search result is satisfactory, the adaptive update processing of the evaluation reference data can be performed by the same procedure. An appropriate method may be used as an algorithm for extracting a common attribute of the plurality of web pages. For example, a method is considered in which a set of words commonly included in a plurality of web pages is regarded as a common attribute.

When the searcher is realized in the form of computer software, it is not always necessary to execute all functions of the searcher on one computer, and various functions of the searcher are distributed among various computers. It may be executed in the form described. Since the search function is realized in a distributed form as described above, even in a case where the dispersed individual partial devices themselves cannot be said to be a search device, it is possible to use the reusable evaluation reference data to select links. Anything that is included in the technical idea of the present invention of searching for a web page is included in the scope of the present invention. In the case of implementing a search device in a distributed form on a network as described above, when it is necessary to transmit evaluation reference data in which private interests are described in detail, in order to protect privacy, evaluation standards are required. It is good to handle the data encrypted.

In addition to the distributed form described above, when searching for a web page that requires a long time, it may not be suitable for displaying a time and a search result list via a network. It is thought that there are many. In such a case, e-mail, FAX, or any other communication means may be appropriately used to transmit the search result to the user.

The present invention can be implemented, if necessary, in the form of a dedicated hardware called a web page search device. In such a case, the effects and advantages according to the present invention can be enjoyed. Therefore, an embodiment in which the web page search method according to the present invention can be realized by a dedicated hardware device is also included in the scope of the present invention.

[0042]

As described above, according to the present invention, attention is paid to the problem of setting conditions in a search, and evaluation reference data that can be finely handled for each search topic is considered and stored in a storage device. This makes it easy to reuse it, and based on this evaluation criterion data, we preferentially load and search for promising web pages as search targets, so detailed conditions that were not possible with conventional search engines It is easy to set up, can search with high efficiency to avoid irrelevant web pages, enables more accurate search compared to conventional search engines, and does not require a dedicated database, Without the hassle of relying on expensive and time-consuming human resources, and is therefore ideal for general users to easily and accurately search web pages. Achieve the results.

[Brief description of the drawings]

FIG. 1 is a functional block diagram illustrating the principle of a web page search method according to the present invention.

FIG. 2 is a flowchart showing an example of a search program according to a search method according to the present invention.

FIG. 3 is a schematic diagram illustrating an example of a method for evaluating a web page and a link based on evaluation reference data.

FIG. 4 is a functional block diagram showing an example of an evaluation reference data adaptation update process.

[Explanation of symbols]

Reference Signs List 10 Information source readable on the Internet or computer storing the information source 20 Search device having search method according to the present invention 21 Evaluation reference data editor 22 Evaluation reference data selector 23 Search condition input device 24 Selection of reading target Unit 25 reader 26 link extraction / addition unit 27 page evaluator 28 search result list creation / display unit 29 evaluation standard data adaptor

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) G06F 13/00 540 G06F 13/00 540E

Claims (11)

[Claims] 1. A reusable evaluation criterion data, which is used as a criterion for evaluating a web page when performing a search on a specific topic in response to a request from a user, is created by the user or a third party. A first step, a second step of specifying evaluation criteria data corresponding to a search topic from reusable evaluation criteria data that has already been created, and execution by inputting search conditions by the user A third step of determining the evaluation reference data at the time; a fourth step of setting a set of web pages waiting to be read; and calculating from the set of web pages waiting to be read based on the evaluation reference data at the time of execution. A fifth step of selecting a web page with the highest evaluation as the next reading target, and the web page selected in the fifth step. A sixth step of reading a web page, and adding a web page referred to by a link in the web page read in the sixth step, which is a future read candidate, to the set of web pages waiting to be read. A web page search method, comprising: a seventh step; and an eighth step of calculating an evaluation of the read page based on the evaluation reference data at the time of execution. 2. The evaluation criterion data includes a list of weighted keywords, and in the eighth step,
The method of claim 1, wherein the evaluation of the web page is calculated such that the evaluation monotonically increases with the number of times the web page includes a keyword and increases monotonically with the weight of the keyword. 3. A step of requesting a simple search service on the Internet such as a search site to perform a search using search conditions having contents similar to the evaluation reference data at the time of execution, and reading the search result. The fourth step is to set a set of web pages listed as search results of the search service as the set of read-ready web pages.
How to search for web pages listed in 4. In the second and third steps,
4. The evaluation reference data according to claim 1, wherein even if the user does not specify the evaluation reference data, evaluation reference data most closely related to the search condition is automatically set based on the search condition specified by the user. 5. How to search web pages. 5. In the fifth step, when a web page belonging to the set waiting to be read is added as a reference destination of a link posted on another web page in the seventh step, The web page search method according to any one of claims 1 to 4, wherein a description of the vicinity of the link on the another web page is analyzed, and the web page is evaluated using the analysis result. 6. The web pages that have been read are displayed to the user as a list in descending order of evaluation, and the user specifies one item of the list by clicking a mouse or the like. The web page search method according to any one of claims 1 to 5, further comprising the step of enabling browsing of the web page. 7. While a process is being performed in a loop formed by the fifth, sixth, and seventh steps, a search end condition is appropriately checked, and if the condition is satisfied, the search is terminated. 7. The web page search method according to claim 1, further comprising a step of performing an end process. 8. The web page search method according to claim 7, further comprising an input device for allowing the user to easily set the search end condition according to claim 7. 9. To cause the user to point out unsatisfied web pages or satisfied web pages from the list of search results, analyze common attributes of those pages, and more accurately reflect the interests of the user. 9. The web page search method according to claim 1, further comprising a step of adaptively updating the evaluation criterion data using the analysis result. 10. A computer-readable storage medium storing a program for causing the computer to execute a web page search method, wherein the program is configured to search for a specific topic in response to a user request. The first step is to create reusable evaluation criteria data, which will be the basis for evaluating a web page when performing a search, and search topics from already created reusable evaluation data. A second step of specifying evaluation criteria data corresponding to the following; a third step of determining evaluation criteria data at the time of execution by the user inputting a search condition; and a second step of setting a set of web pages waiting to be read. And calculating from the set of web pages waiting to be read based on the evaluation criteria data at the time of execution. A fifth step of selecting the web page with the highest evaluation obtained as the next read target, a sixth step of reading the web page selected in the fifth step, and a read step of the sixth step A seventh step of adding, to the set of web pages waiting to be read, web pages that are future read candidates among web pages referred to by links in the web page, and evaluating the read pages at the time of execution. An eighth step of calculating based on the evaluation reference data. 11. A system comprising a plurality of devices which are physically distributed, wherein each device has a partial function for executing a web page search method as an entire system. These partial functions collectively comprise a first step of creating reusable evaluation criteria data that serves as a basis for evaluating a web page when performing a search on a particular topic in response to a user request. A second step of specifying evaluation criteria data corresponding to a search topic from among reusable evaluation criteria data already created, and an evaluation criteria at the time of execution by the user inputting search conditions. A third step of determining data, a fourth step of setting a set of web pages waiting to be read, and from the set of web pages waiting to be read. A fifth step of selecting a web page having the highest evaluation calculated based on the evaluation criterion data at the time of execution as a next reading target, and a sixth step of reading the web page selected in the fifth step. And a seventh step of adding, to the set of web pages waiting to be read, a web page referred to by a link in the web page read in the sixth step that is a candidate for future reading, An eighth step of calculating an evaluation of the read page based on the evaluation reference data at the time of execution.