TWI391834B - Systems for and methods of finding relevant documents by analyzing tags - Google Patents

Description

System and method for finding relevant documents by analyzing tags [related application]

The present application is based on US Provisional Patent Application No. 60/705,704, filed on Apr. 3, 2005, and the title of "Technology for Searching for Relevant Documents Using Label Analysis", in accordance with 35 USC § 119(e) The priority of the case, which is easy to incorporate as a reference.

The invention relates to the search of documents, and more particularly to the system and method for searching for information on the Internet by analyzing tags generated by users, thereby improving the quality of search results, or Correlation.

The invention relates to the search of documents, and more particularly to the system and method for searching for information on the Internet by analyzing tags generated by users, thereby improving the quality of search results, or Correlation.

The Internet search engine is designed to locate the information you need from the vast expanse of information that is included throughout the Internet. First, the user describes the information he or she is looking for by entering a request to include the search item, and the search engine is associated with the web pages that are used to identify the most likely to use the information sought. The calculation of the target, and the search item is compared with the index of the webpage, and then the search engine will reply to the sorted list of hyperlinks pointing to the webpages, and the links of the webpages considered to be the most relevant will be Listed closer to the top of the list.

The goal of the search engine is to deliver the most relevant pages to a request. Search engines use a variety of techniques to determine the relevance of a web page, for example, by considering the information contained in each page, for example, the existence, density, of such search items in the document. And proximity, by considering information about hyperlinks between such web pages, or by user behavior, such as clicking, browsing, or evaluating results or web pages. In order to achieve the best results, these techniques can be applied separately or in various combinations.

Since the number of web pages on the Internet is very large and growing, plus a large number of web pages that meet the requirements of users on the surface, it is decided which pages are most relevant. The procedure is quite difficult. Similarly, since most users are not proficient in the requirements of the generation and input architecture, it is possible that the type of information they are looking for becomes meaningless. The situation occurs, and therefore, by making a comparison between the words in the requirements and the text in the documents to determine which documents are most relevant to the request, the correctness of the results that can be provided is rather limited.

When users browse or search the Internet, they can target various purposes, such as web pages, images, topics, weblogs (also known as blogs), or other targets. , by adding a reference to the target, adding it to a "bookmark", wherein the bookmarks may contain one or more "tags", and the label is composed of users One or more items associated with the target, a hyperlink to the target (uniform resource locator, or "URL"), a mechanism for recording the relationship, and potentially other information And, these bookmarks will help the user to call the target and any tags to assist in re-calling or generating connectivity to other targets related to the bookmarked target, for example, if a user visits When describing a web page relating to a solar panel for a roof, it can bookmark it and use the item "solar" to generate a label associated with the web page. In addition, it can also utilize the Another page of the state solar power rebate program is associated with the same label, so the result will be that the label with the item "solar" will be linked to two pages. All have an association.

There are many ways for users to enter tags. For example, use a server application, an applet in the bookmarks toolbar, a browser plug-in or an extension. , a client application, or some other application, wherein once the tag is entered, the user is typically allowed to search for the tags to display the pages associated with the tags And, by now, services have been provided that allow users to search for tags that they use themselves, or to search for tags for others.

Bookmarks provide a hint that a user cherishes a goal, such as a webpage, and the tag additionally provides a hint that the user associates the goal with a certain item, or certain items. And this information is potentially valuable in deciding whether a web page should be displayed as if it were from one of the search engines, because it not only implies the real user’s interest in the web page, but also implies For a particular topic of association.

Therefore, there is a need for a search engine that will be relevant to various web pages, images, blogs, or other purposes when deciding which pages, images, blogs, or other goals are relevant to the user. The label is considered.

Embodiments of the present invention provide a user-target list (the result list) in response to a search request, wherein the result list is organized based on a relevance of the request to each target, and is preferably The relevance is to add the targets to the tag, bookmark the targets, or both, or to indicate the relevance or value of a target to any other user of the search. basis.

In a first aspect of the present invention, a method of determining a relevance of a plurality of targets to a search request includes recording "bookmarks, and/or" directed to the plurality of targets such that the plurality of tags and the plurality of tags A plurality of goals are associated, and a correlation is determined between each of the plurality of requirements and any given requirements. The method is used to organize the target later to display within a range of results in response to a search request, and the target includes, pointing to a web page, text, image, photo, label, tag group , hyperlinks, or hyperlinks to a topic area, concept, user profiles, answers, audio files, video files, software, or any combination of these.

Each of the plurality of tags from the plurality of tags will contain one or more items, and the method further includes associating each item from the one or more items with a target to thereby define a , or a plurality of corresponding project-target pairs, and a project score for each project-target pair, thereby indicating a degree of correlation between the project and the target. Alternatively, or additionally, the method also includes bookmarking the target.

Preferably, the achievement of the determination of a relevance score for a tag-target pair is achieved by combining the item scores for the item-target pairs for each item in the tag, Moreover, the combination of item scores is achieved by summing them, or by weighting each item score by using a weight and summing up the weighted item scores.

In one embodiment, a relevance score for a tag-target pair is determined from the number of occurrences of an item in a tag that has been associated with the target, and the number of tags associated with the target, The number of times the tag has been associated with the multiple targets, or any combination of these. In addition, a correlation score for a tag-target pair may also be determined from the number of tag-target pairs containing one of the items in the tag, including the number of tag-target pairs associated with one of the targets, or Both.

In another embodiment, the method may include: causing a tag to be associated with a target from the plurality of targets by a first user, and performing an inclusion by a second user Searching for one or more items in the tag, organizing the plurality of targets in the list of results based on the relevance scores, thereby defining an organized list of results, and The organized result list is replied to the second user, wherein a relevance score for a target and the search request corresponds to each item of the search request that appears in the target or is associated with the target a relevance score, and, alternatively, or additionally, a relevance score from the plurality of tag-target pairs of the tag-target pairs is determined from the first user having The number of any associated tags of the plurality of targets, the number of targets that the first and second users have associated with the tags, or both.

In another embodiment, a relevance score for a tag-target pair is determined from a confidence assessment of the first user, the second messenger, or both, and a confidence The evaluation is a measure of the similarity between the bookmarked, generated label, or search activity of the first and second users by the selected user having made an assessment of the label associated with the target. A similarity metric, a relationship metric between the first and the second ambassador, or any combination of these.

Organizing the plurality of goals may be accomplished by sorting the plurality of targets based on the relevance scores (eg, the highest ranked target is listed in the first), or by utilizing one The graphical element marks at least one of the plurality of targets.

A tag, a bookmark, or an evaluation system can be used to input the tag into a field presented to a user, evaluate the tag, and block a link to the target (by thereby marking a "negative" "Associate", select the tag, select the target, detect a bookmark, or use the tag to perform a search for the target to generate an association. In one embodiment, a tag crawler is to associate at least one of the plurality of tags with at least one of the plurality of targets.

In a second aspect of the present invention, a method for establishing a system for replying to a target that has been organized in a result list includes storing a plurality of targets in a tag database Correlating a plurality of tags, and storing a correlation score between the plurality of tags and the plurality of targets in an index database, wherein the relevance scores are used in an organized Organize multiple files in the results list.

A plurality of tags are stored in a tag database by storing items from the plurality of tags, and the relevance scores indicate a correlation between the items and the targets, and The method also includes storing a plurality of indexes in the index database, and each index input corresponds to an item from the plurality of items, a corresponding target from the plurality of targets, and in the item and A corresponding relevance score between the goals.

In one embodiment, each of the corresponding relevance scores between a project and a goal is related to a confidence of the user who associates the project with the goal, and alternatively, Or additionally, each relevance score for a project and a goal is determined by the number of times the goal has been bookmarked, or the number and value of the assessment that the goal has been provided, and One, or additionally, a relevance score between a project and a goal is determined from a statistical taxonomy, or a ranking regression algorithm, for example, logistic regression, support vector machines (support vector machines), classification or regression trees, or boosted tree ensembles.

The method also includes presenting a list of results to a user in response to a search request including a target by which the item is associated with a target included in the list of results, and Determining a relevance score between the item and the goal, wherein a relevance score between a goal from the plurality of goals and a tag associated therewith is determined from the tag has been generated with the goal The number of associations, the total number of labels that have been associated with the target, the number of times the label has been associated with any of the multiple targets, and the number of labels that have been associated with all of the multiple targets, the label The time that has been associated with the goal, or any combination of these.

In a third aspect of the present invention, a method for organizing a plurality of objects for display in a result list includes causing an item in a search request to be associated with a tag associated with a plurality of targets And replying to a list of results that have been included in the plurality of goals organized on the basis of the associations.

In a fourth aspect of the present invention, a system for replying to a search result list in response to a search request includes a tag database for storing tags associated with the target, and a tag analyzer coupled Access to the target database, wherein the targets include pointing to web pages, hyperlink groups, text, images, photos, tags, tag groups, subject areas, concepts, user profiles, answers , hyperlinks to audio files, video files, software, or any combination of these, and preferably, such targets are hyperlinks to web pages.

The tag analyzer is programmed to determine a relevance score between a tag and a target, and in an embodiment, the system also includes a target index for storing the tag and the target Correlation score.

In one embodiment, a relevance score is determined by summing the items used to form a label and the weighted relevance score of the target. In another embodiment, a relevance score between a search request including a project and a goal is determined by the number of tags included in the search request, and is included in the search request. The number of times a tag is included in the tag database, the number of tags that have been associated with the target, the number of items in the tag and the items in the search request, or any combination of these . In still another embodiment, a tag and a relevance score of a target are based on a position of the tag within the target range, a frequency of the tag within the target range, and the tag is at the target A density within the range, or any combination of these.

In an embodiment, the weighting of a relevance score for a tag and a target is based on a degree of confidence (confidence assessment) assigned to a user who associates the tag with the target, and, The determination of the relevance score is determined by using a statistical classification, or a sorting regression algorithm, a clustering analysis algorithm (cluStering analysis algorithm), or a morphological analysis algorithm. The statistical classification, or sorting regression algorithm, may include logistic regression, support vector machines, classification trees, or classification tree ensembles.

In another embodiment, the system also includes a search engine coupled to the target index, wherein the search engine is programmed to receive search requests including items corresponding to the tags, and to tag- The target pair's relevance score is used as a basis for replying to an organized list of results. In addition, the system will also include a user database coupled to the search engine, and the user database will include relevant search requirements. Information, for example, a link tracked by a user, a tag associated with the target, a target blocked by the user, a bookmark, or any combination of these.

Embodiments of the present invention, unlike conventional search engines, utilize tags, and/or bookmarks to provide more relevant information to users searching the Internet. In one embodiment, a search engine executes an established method to receive a request and determine a list of related files, or groups of files, to analyze tags associated with the file, or group of files, to determine a A list of files, or groups of files, and a combination of the two lists in some manner to deliver to the user a list of results. Here, it can be understood that when the following examples describe the search and reply files, the present invention can be used to search and reply to any target, including, but not limited to, for web pages, text, images, photos. , hyperlinks, hyperlinks, audio files, video files, software, or any combination of the above, hyperlinks, or hyperlinks, are used to list a few targets.

For example, in accordance with the present invention, a request with item X would result in a list of web page results, including a web page M in a certain portion of the list, then a first user would utilize a project containing the item X. The tag is associated with the web page M, and a second user who performs a search using the item X in a request is delivered a result list, and the result list displays the web page M in a ratio It is a higher position that has been displayed before the first user created the tag.

In accordance with an embodiment of the present invention, determining the extent to which the web page M is considered to be more relevant to the second user is by analyzing factors including, but not limited to, the item X being used in a tag for the web page M. The number of times, the total number of all the tags associated with the web page M, the number of tags that the first user has created, the number of files each user has added to the bookmark or tag, and the item X is used from start to finish. The frequency of a tag, the total number of tags from beginning to end, including the number of tags/file pairs of the item X, for the page M, the relationship between the first and second person, or both Reference, the degree of confidence of the first and second person, or the group to which the first person or the second person belongs. In addition, similarly, the relationship, or degree, between the users and the groups to which they belong, as well as the degree of confidence of the users, can also be quantified using metrics, for example, two users. When the metric is expressed as 1, it means that they will be more similar than the two users with the relationship metric of 0.5 (for example, have similar interests, or share a common friend). In addition, if there is more than one word in the item X, other factors can be analyzed, for example, the number of words of the item X included in the request of the second user, and whether the words are It is used as a phrase, the order of the words, and all the factors mentioned previously, and the analysis can also include analysis of different combinations of words.

In accordance with other embodiments of the present invention, the search engine may be unable to deliver a list of results for different rankings, but rather some graphical elements may be set to indicate which files have been included because they have tags associated with them.

According to other embodiments of the present invention, the tag may not be explicitly associated with the file by the user, but may be by checking the bookmark, guiding the search, or other user's behavior, such as rating, Block, save, click.

In accordance with other embodiments of the present invention, the tags may be related to not only web pages, or groups of web pages, but may also be associated with any identifiable source of information, public or non-public, including, but not limited to, images, photos, other tags. , tag group, subject area, user profile, concept, map, audio or video file, software, or other target.

In the following full description, the term "search engine" is used to mean a device (or a program that runs on a general purpose computer) that is used to input a request and generate an electronic a list of results for a hyperlink to a document, or web page, or other target that is accessible over the network, and the search engine will include an index of the files in its collection to determine the relevance of each file. Encoding and algorithms, and a graphical user interface that delivers the list of results to the user.

In the following full narrative, the term "query" refers to a set of items submitted to the search engine, whether typing, voice input, or through a "link" that has been embedded in a set of search items. Submit it or submit it through any other interface. In addition, a requirement system may include a single word, a plurality of words, or a phrase, and the requirement may be expressed in words as a question (for example, a "natural language requirement"), an imprecise A portfolio of projects, or a constructed Boolean representation. Even a request system can include a search engine for searching for electronic files containing or related to search characters, or for symbols of web pages, or any other character.

In the following full description, the term "website" is used to mean a collection of web pages that are linked together and available in the World Wide Web, and the term "webpage" is used. It is a file that is displayed on a website and accessible from any number of hosts through the World Wide Web, including, but not limited to, text, video, video, music, and graphics.

In the following full description, the term "results list" means a hyperlink or a list of hyperlinked groups, which can utilize Hypertext Transfer Protocol (HTTP), or Is a reference file, target (as defined above, including but not limited to, video and video), or any other web page used for accessing any other communication protocol used to access web pages or other electronic files. Information about a link, including, but not limited to, the title of the documents, a summary of the documents, the number of associated tags or other related metrics, a list of associated tags, and a buffered copy of the files. The link, the date on which the documents were last indexed, or the date of the latest revision, the images associated with, or located in, the information retrieved from those documents, and the users who can bookmark or tag them. .

In the following full description, the term 'tag' is used to mean any data structure containing any one of: one or more items, each of one or more words, with reference to an addressable target Hyperlink, and other information, such as when the tag was generated and the user who generated it. A tag can contain a link to multiple targets, for example, a web page, an image, a map, or It's the other target on a computer network, whether it's on the Internet or on a regional computer storage device. In addition, "tagging" also means pointing an item and a point to an addressable file. , or a special hyperlink to the target that produces the associated program.

As used herein, the item "bookmark" is used to record any data structure of any hyperlink, an indication of the user who created the bookmark, the time the bookmark was created, and a time as defined above. label.

As used herein, the term "document" is defined broadly and, in addition to its original meaning, includes computer files and web pages. The term "file" is not restricted to computer files containing text, but may also include user files, concepts, answers, computer files containing graphics, audio, video, and other multimedia materials. A user profile is a page or record that contains, but is not limited to, information about a person, such as their interests, habits, friends list, face, career experience, and education, which are some of the information used for example.

As used herein, the term "spammer" is defined as any number of techniques that are designed to capture the advantages of the relevant methodologies of the search engine, while attempting to make a search engine A person or entity that displays a higher rank or more frequency associated with a link to its product, web page, or other material than the search engine uses a different method.

As used herein, the term "programming" means a hardware, software, firmware, or other tool used to perform computer instructions for storing, processing, transmitting, or otherwise manipulating data. Any combination.

As will be described in greater detail below, a search engine will accept the requirements entered by a user and will utilize various correlations with the purpose of identifying those documents most likely to be relevant to the information sought by such users. a calculation to compare the search terms with a file index, and then the search engine generates a sorted list of hyperlinks to the files, wherein the most relevant files are considered to be the closest Above the list. In accordance with the present invention, a user has the ability to generate a tag that associates an item with a file, and a search engine replies to a list of results based, at least in part, on the analysis of the tags associated with the web page.

In accordance with the present invention, the extent to which a tag affects the relevance of a document to a given requirement may be related to the degree of confidence of the user who created the association, and the degree of confidence may be determined by a number of factors. For example, how relevant is the user's tag in the past, and how similar the activities observed by the user are to other users of the hobby, or to the user performing the request, and the degree of connection between the users, And other factors.

1 is a screen shot of a graphical user interface (GUI) for displaying a result page 100 obtained in response to a request in accordance with the present invention. The list of results can be reordered or flagged based on an analysis of the tags associated with each link.

The result page 100 includes a space 110 for inputting a required item, and an area 120 including the result list replied by the search engine, and the area 120 may also include an replies associated with the search engine. A list of tags 150 for each result. As will be described in greater detail below, in a preferred embodiment, some or all of the results of the region 120 have been reordered based on the analysis of the tags 150. Or, in another embodiment, the results in the region 120 are also reorderable, but some of them are labeled with the graphical component 140 to indicate for the tags, and/or The analysis of bookmarks 150 will have an impact on their relevance, and in addition, the results page 100 will include a mechanism 165 for evaluating the goals.

The user can establish a correlation between the files and the items they think can describe the files, and the program, as previously described, is referred to as "booking" or "generating" Tagging, the completion of the procedure is by clicking on a hyperlink in area 120 or pattern element 160 in the example of adding a bookmark to initiate a record for later reply. Mechanism, either by using an extension program in the bookmark toolbar, or a toolbar, or an applet, or, in the example of generating a label, generating a new label, or multiple new labels and a file The association, wherein the component 160 can be a text link, an image, for example, a disc, or any other image that would suggest "booking" or "generating a label" of the file. Since different users will have different ideas about which projects are associated with different files, it will be possible to create a collection of tags with a wide variety and variety. This set of tags is the analyst in accordance with the present invention.

The user can block the link to a file from a list of results. If they do not consider the file to be related to the request, the program is called "blocking" and the program is By clicking on a hyperlink in the area 120, or the graphic element 170, a mechanism for blocking the connection to a file is initiated, wherein the element 170 can be a text link, an image (eg, the first Figure, 170), or any other image that would suggest "blocking" or "removing" the file. Blocking the file will have a result associated with a negative tag for the file containing the item in the request, and since different users will have different items associated with each other. Different ideas of documents sometimes have different opinions about whether a label is suitable for a document, but at other times, spammers will deliberately make error-oriented labels and documents. Generate an association. And as explained in more detail below, positive and negative labels can be used to compensate for different opinions and to reduce the amount of spam or other misdirected files.

The results page 100 can include an area 180 for displaying a narrative relating to a concept related to the required item, and an area 190 containing "for other concepts related to other required items" (see also See also)" link.

As shown in the embodiment of FIG. 1, when a user inputs the required item "U2" in the space 110 and requests a search, the result page 100 is returned to him, wherein This area 120 will contain a list of results linked to the links associated with the target of the required item "U2". In a preferred embodiment, the results 130 are reordered based on an analysis of the tags 150 that the user has associated with the various files, for example, at www. The file titled "U2 Home: @U2..." by atu2.com has been associated with the items "U2", "U2 Fan Station", and "U2 Fans", as shown in the Tag List 150. The analysis of the label has caused the file to be listed in the higher position of the result list, compared to other methods. In another embodiment, the results may be reordered and labeled, but some of the results will be marked with a graphical element 140 to indicate that the user tag, and/or the bookmark analysis indicates It has been found that the results are more relevant and, optionally, the number of people who find their relevance. For example, the file titled "U2 Platform" at www.u2station.com is already The tag is associated with the item "U2" and the "U2 fan station", and the graphical element has been marked with a humanoid 140 to indicate that other users have found their relevance. Here, it can be understood that a graphic element other than a human figure is also used to inform the user that the relevance has been instructed by other users.

If the user wishes to bookmark a file and/or generate a label, for example, www.u2log.com , the user can choose to click on the graphic symbol 160, and the graphic symbol can be activated. a mechanism for adding a bookmark, and/or adding a label, and the added label can be the same as the existing label, the search item in the area 110, or some other meaning to the user. project. If the user has completed a different search, for example, "Lyrics", and the user does not believe that the file titled "U2 Wanderer.org U2 Album and U2 Lyrics Station" should be listed. When the result list of the "Lyrics" is requested, the user can block the search result having the result that the label containing the item "-lyrics" will be associated with the file by clicking on the element 170. The file obtained in the text, wherein the minus sign ("-") indicates a dissenting, not affirming, association between the item and the web page.

Continuing with the example, the region 180 includes a concept describing the orchestra "U2" and a list of other concepts related to the project "U2", the region 190 including a "see also" connection set pointing to the related topic. Lift For example, "Bono, U2 concert, most sing-along artist, unified action, on-site...", according to the present invention, if the user chooses one of these links, for example, "the scene At the time, a request will be directed to use the search item "sites" to generate a result page similar to the result page 100, wherein the search item 110 is "site eight" and the result list 120 is a list of files associated with the search item, and the locations in the list are sequentially affected by the tags associated with the files, so the order from any search results is subject to the tag Impact. The subject area and concept are filed on February 27, 2006, entitled "Methods of and Systems for Searching by Incorporating User-Entered Information", in US Patent Application No. 11/364,617 As described, this system is also incorporated by reference in this case.

It will be appreciated that many modifications are possible in accordance with the invention. For example, the label generated by the user can be read from a file or imported from another service, in addition to being directly input from a terminal by the user, and when the result page 100 is displayed, When the result list 120 and the tag list 150, as well as the concept 180, and the link to the concept 190, it will be understood that, in accordance with the present invention, the resulting page affected by the tag analysis can utilize the area, including in Figure 1. The displayed ones, or those other than those, or any combination of some of these areas are displayed, and the label information is used to combine with various page design elements. To produce a broader, correct, and more meaningful search result.

2 is a flow diagram in accordance with the present invention for illustrating the operation of an internet search application 200. The Internet search application 200 provides the ability for a user to submit a request to the search engine and receive a result determined at least in part by a tag analysis to thereby provide the user with a provider other than the method provided. For more relevant search results, the user can browse the web pages displayed in the results list, and they can also choose to "book" some of the pages to indicate that they are discovering the pages. Relevant, or irrelevant, in addition, they can use the request item, or use some other items, or multiple items to generate the labels of the pages, and the search engine will record any submitted items. Labels and use them in future searches performed by other users.

In step 210, the user submits a request to a search engine, and then the program proceeds to step 220 and step 230. In step 220, the search engine utilizes various uses. The information recovery method of each type of algorithm compares the requirements with the target in the corpus to assemble a list of the most relevant files, and in step 230, the search engine analyzes the various files. Such tags are used to assemble a list of the most relevant documents, wherein the analysis of the tags may be the same for all users, or may be tailored to the individual user performing the search. Or tailored for the group in which the user is one of the members, then steps 220 and 230 continue to step 240, and in step 240, from step 230 The results are combined with the results from step 220 to provide more relevant results, and then the program continues to step 250, and in step 250, the results page (eg, 100, 1st) ) To be sent to the user system. Further, the user can select from the Department of step 250 proceeds to step 260, or 270.

In step 260, the user will follow one or more of the links to view the files in the results list, or, alternatively, in step 270, The user enters a bookmark and, optionally, enters a label, and each of the actions associates one or more items with one of the files in the list of results, so In order to enter a tag, a file can be clicked on a mechanism that provides a user interface for the user to enter the tag to bookmark a file or label (eg, area 160, Figure 1). Or, in step 260, the user can use a "bookmarklet" contained in his browser, or other similar mechanism to bookmark the file and enter the tag, or, Alternatively, in step 270, the user can block files deemed to be unrelated to the request by clicking on a mechanism for blocking. Thus, from step 260, the file in the results list is visited, the user can proceed to step 270, enter the tag, and, alternatively, the user can proceed from step 270 to step 260, and Both steps 260 and 270 are passed to step 280, in which the system records the bookmarks, tags, and evaluations entered by the user, and then step 280 returns to step 230, now, The database of tags that will be analyzed during any subsequent searches will include the new tags entered in step 270, and in addition, parallel to the loop back to step 230, the process may continue to step 290. In step 290, the user will conclude his search.

Figure 3 is a diagram illustrating the components of a system 300 in accordance with the present invention. The system 300 includes a user client 305 coupled to a network server 310, and the network server 310 is coupled to a search engine 320, a user database 330, and a tag. The database 340 is coupled to a file index 350. The user database 330 is also coupled to the search engine 320. The tag database is coupled to a tag analyzer 360. And a tag crawler 391, and the tag classifier 360 is also coupled to the file index 350, and the file index 350 is coupled to an indexer 370 in turn. The indexer 370 is coupled to a web content database 380, which is coupled to a web crawler 390, and the web crawler 390 and The tag crawler 391 is coupled to one or more websites 399 via the Internet 395.

In operation, the web crawler 390 navigates through the Internet 395, visits the website 399, and creates the web content repository 380 using the content of the web pages it accesses, and The indexer 370 uses the network content database 380 to generate the file index 350. The tag crawler 391 navigates through the Internet 395, visits the website 399, and utilizes its location. The tag database 340 is created by the tags found.

When a user directs a search, it uses the user client 305 to enter a request and is presented to the web server 310, which then submits the request to the web server 310. Search engine 320, which then uses the correlation algorithm and the factors derived from the tag analysis described above to compare the request with the file index 350 to determine the most relevant file, and will The list of results is replied to the web server 310, and the web server 310 then delivers the resulting page (e.g., 100, Figure 1) to the user client 305 for display.

And, in response to the request, the user database 330 will record information about the user's search, for example, a link from the result list (eg, area 120, Figure 1), bookmarked, or already A file for evaluation (eg, step 165), and a tag entered using the tag input mechanism (eg, region 160, FIG. 1) and the blocking mechanism utilizing the effect of having an input negative tag (eg, region 170, 1) the blocked file, which can be used by the web server 310 and the search engine 320 to determine the subsequent search results for the user and to determine the label to be used in the user. How much confidence is placed therein, and in response to a request, the user enters the tag entered by the tag input mechanism (eg, region 160, FIG. 1) with the user using the blocking mechanism (eg, region 170, The negative label entered in Fig. 1) is also recorded in the label database 340. In an embodiment of the invention, the information stored in the user database 330 and the tag database 340 can be implemented as two separate databases, or they can be implemented as the same data. In the library.

Under some immediate basis, but not necessarily, when required to be implemented, the tag information contained in the tag database 340 is sent to the tag analyzer 360, where An analysis is performed to determine which impact on relevance is to be determined by the various tags associated with each of the files to determine the most relevant web page for the request by the search engine 320. Tag analyzer 360 will record this tag relevance information in the file index 350 for subsequent searches.

The tag database 340 sends features to the tag analyzer 360, including, but not limited to, request items, user identifiers, file IDs, file links, tag items, rating levels, and time stamps ( Time stamps), in addition, the tag analyzer 360 can also find other features for the given file, including, but not limited to, the density of the items within the scope of the file, the items within the scope of the file Locations, the existence of such items in various paragraphs of the document, and hyperlinks to the documents containing the items, in addition, the tag analyzer 360 can also find other features for the given file, including, But not limited to, previous tag history, bookmark history, confidence level, similarity to other tags (eg, similarity between the search items used and the tags created by this user and other users) , and members in the group.

The tag analyzer 360 uses these features to develop a set of relevance scores for various files based on different tags, wherein the program by which these features are analyzed is shown in Figure 4, and the mechanism is It is used to calculate the degree of a user, or the relevance of a generic solution, for any given request-file pair.

An analysis is performed on the label material previously discussed. In general, the relevance of any document provided will be a function of the factor for any given requirement, including, but not limited to, the number of labels containing the items in the request, any labels provided. The number of times used in the corpus of tags, the total number of tags for the document given, the number of similar tag document pairs, the words in the matching item The number of times the file has been bookmarked, the number and number of ratings for that file. In addition, for a document provided, the ability of any label to predict relevance is proportional to the level of confidence of the user entering the label, if it can be evaluated. Here, it can be understood that the correlation modeling program according to the present invention can be implemented by other forms of analysis as well as other methods, including, but not limited to, any statistical classification, or sorting regression algorithm, for example, Logistic regression, support vector machines, classification or regression trees, or boosted tree ensembles.

Figure 4 is a flow diagram in accordance with the present invention to illustrate the step 400 for preparing and analyzing tag data as a preparation for determining the relevance of a file corresponding to a request.

Referring to FIG. 4, in step 410, the tag data is input by the user through a web client and a web server, or is crawled by a tag. A tag crawler (eg, 391, Figure 3) is entered and, by the system (eg, step 280, Figure 2), is presented to a tag repository (eg, 340, 3)), wherein the step 410 can continue continuously regardless of the remaining steps of the program described in FIG. 4 for a period of time.

In step 420, each of the tags in the tag data (340, Figure 3) is analyzed, and then the process of analyzing each tag proceeds to step 430, calculating user confidence, and steps. 440, determining the number of weighted labels. In step 430, for each input tag, the degree of confidence of the user who inputs the tag is calculated, and the degree of influence of the tag on the relevance of the associated file is A function of the degree of confidence of the user who entered the input.

The degree of confidence can be calculated using an algorithm or using a statistical model of user behavior, including, but not limited to, how much the user's behavior is consistent with the user community. The behavior of the item, or subject area, as a basis, or because other users connect to the social network, join, or identify (subscribing) To) the user enters the tag to base the user with confidence, add bookmarks, generate tags, click, sort, or block. For example, if a user, Luke, uses item A to generate a label for the given file X, and at the same time other users have used items D and F as the label of the file X, then The complete set of tags for file X would be {a,d,f}, where the lowercase symbol "a" represents a real condition of using the item "A" as a label, continuing this example, if the other two users , Simon and Peter, using the request item A to search, and each of them blocks the file X, now the file X will be labeled {a,-a,-a,d,f}, So, the result will be that Luke's level of confidence will be reduced by the fact that multiple users disagree with his tags, and Simon and Peter's confidence will increase because their tags are consistent with multiple users. Here, it can be understood that there are other methods consistent with the present invention that can be used to determine the degree of confidence of a user, but if the user is unknown or the degree of confidence of the user cannot be determined, then The level of confidence that will be neutral or taken from the source of the label is assigned to the label. In addition, it will be understood that user confidence can be analyzed at the time of the label or through some other The program that takes place at the time is calculated.

In step 440, the number of weighted labels for each document or group of files is determined for each label. If a file X has been added to the label n times using the item A, then the file The number of the weighted labels of X for the item A will be, as a factor, the degree of confidence of each user U _i that generates each of the labels a _i , i=1 to n, among them, as the file X All the labels a ₁ to a _n of the reference, whether it is a positive or negative aggregation, in addition, if a user inputs a lot of labels, the user can be regarded as frequently performing the file The label, and the importance of any label given by the user, is considered to be lighter than the label of the user who made the label less frequently, and the label is earlier than when it was created. The lower is considered to be less important, and if they are created at a point in time closer to the current time, the importance is relatively heavier. Here, it can be understood that other factors that can be considered in determining the number of weighted tags will also be considered as part of the present invention, and once each document is determined for the number of weighted tags for each tag. After that, the entire process will proceed to step 450.

In step 450, each item in the corpus is analyzed to establish a tag score for each document, or group of files, for each item, and then analyze each item. The process continues to step 460 to analyze each document and step 470 to calculate a label score for each document.

In step 460, the next file to be subjected to the calculation of the tag score is analyzed, and all the information previously collected or calculated about the file is first gathered together, and then an additional analysis is performed. For example, factors that can be considered include, but are not limited to, the occurrence, or density, of the item in the document, the location of the item in the document, and multiple items in the same label. The existence of the item, the presence of the item in the link to the hyperlink to the file, the time the label was created, or the latest revision, and the similarity of the item to other items in the document. This can be based on statistical analysis, clustering analysis, or morphological analysis, or other forms of analysis that determine similarity.

In step 470, the current file's score for the current tag is calculated. The label score for each file is a function of the total number of labels pointing to the document, where each label contributes to the calculation based on the weight applied to it, which is the number of weighted labels, just as Determined in the above step 440. Moreover, the contribution of each tag to the tag score is proportional to, as determined in step 430, the confidence assessed by the user entering the tag, and the file completed in step 460. Analysis, in addition, the label score S _{A of} the document X for the item A will be the total number of different items appearing in the label database (for example, 340, Fig. 3), the item A appears in The frequency in the tag database, the number of different items by which the document X is added, the total number of items X associated with the document, and the number of different files that are added to the tag using the item A, and The combination of these factors is calculated so that the label score for each document is assigned a value, and it is understood that other factors that can be calculated can be considered as part of the present invention. . In a preferred embodiment of the present invention, the tag score is tailored to the individual user or group of users, and in another embodiment of the present invention, The tag score for each document is the same for each user of the system, and then step 470 will return to step 450 until each of the files in the collection has been A label score for each item assigned to it, and then the program proceeds to step 480.

In step 480, each document that has been tagged by a given tag is indexed, so that the tag score for each item for each item is recorded as being searchable. A simple and fast search format to determine the relevance of all documents related to the required item, as determined by label analysis, in addition to the index of the document and its label scores, label weights, and user confidence The degree can be published as a file index (for example, 340, Figure 3) that can be quickly and easily searched by a search engine to calculate by tag analysis alone or in combination with other search technologies. The relevance of each document that is determined for the required project. Here, it can be understood that in other embodiments of the present invention, some steps of the program 400 may be omitted, or other steps may be inserted, or different weights may be applied, or It is still considered to fall within the scope of the present invention to calculate different label scores.

Figure 5 is a flow diagram illustrating a routine 500 for calculating results using tag data in accordance with an embodiment of the present invention.

Referring to FIG. 5, in step 510, the search engine processes the request (eg, 230, FIG. 2) that may include one or more items.

In step 520, the search engine generates, based on each of the requirements, a list of files or groups of files most relevant to the request, and then based on each item. The program for determining the correlation will proceed to step 530 to identify the file, 540, to determine the label ordering, and 550, to score each file. In step 530, in order to be in each of the requirements, a list of such documents deemed relevant is generated based on the tags on which the files have been associated, and The list can be from very short (for example, 5 or less) to very very long (for example, 10,000,000, or more) in length, and it can be understood that this list can be used for searching applications. The actual purpose is truncated and, depending on the needs of a particular embodiment, may be classified or not.

In step 540, the label score for each document is determined with respect to an item, or any group of items, and the label score for each document will be assigned to a function of the tag score of the file in the index, and the extent of the current confidence of the users who will be presented with the tags for calculating the tag score, and may be for individual users Or the users of members in some groups vary.

In step 550, based on the request item currently under consideration, each document may be scored using a value that determines which position in the result list the document is to be placed in a result list, after which Step 550 will return to step 520 until all items in the request are considered.

In step 560, the relevance scores of each of the documents based on the items in the request are combined to calculate the overall requirements of each document relative to the complete requirements presented. Correlation scores, next, in step 570, a sorted result list is generated, and, in step 580, the result list is delivered to the search engine, regardless of what is used. Other correlation methods are combined (e.g., step 240, FIG. 2). It will be understood that in other embodiments of the invention, certain steps of the program 500 may be omitted or may be Processing in a different order, or other steps may be inserted, or different weights may be applied, or different label scores may be calculated, and are still considered to be within the scope of the present invention.

Figure 6 illustrates the hardware components of an application system 600 searched by an Internet using a user 610 in accordance with the present invention. The system 600 includes a client device 620 coupled to a network server 640 via the Internet 630, and the client device 620 can be any device for accessing the network server 640. And, the architecture is to utilize the Internet Protocol, including, but not limited to, HTTP (the Hypertext Transfer Protocol) and WAP (Wireless Application Protocol). Preferably, the client device 620 is a personal computer, but it can be other devices, including, but not limited to, a handheld device, such as a mobile phone or a personal digital PDA. And can utilize standards such as HTML (Hypertext Markup Language), HDML (Handheld Device Markup Language), WML (Wireless Markup Language), or the like. And present information.

The network server 640 is coupled to a search engine 650 and a tag data store 660. The tag data store 660 is coupled to a tag analysis server 670, and the search server 650 is coupled. To the index data store 680, additionally, the tag analysis server 670 is coupled to the index data store 680.

Figure 7 shows a file index 700 in accordance with an embodiment of the present invention. Those skilled in the art will appreciate that the file index 700 is a conceptual construct for explaining the methods of the present invention, and that preferred file indexing will use inverted indexes. The file index 700 includes exemplary first and second columns 740 and 750, and each of the columns includes a tag-target pair and related information in columns 705, 710, 715, 720, and 725, respectively, see the column. 740, wherein the column 705 includes the tag "U2", the column 710 includes a target, here a hyperlink to a web page ("U2 Home"), the column 715 includes a pair of the tag-target pair ("U2-U2 Home") column (eg, unweighted) correlation score (95), the column 720 contains a weight for this tag-target pair, and the column 725 includes for the tag U2 A user confidence assessment associated with the user of the target "U2 Home." The column 750 also contains similar identification information. The input (0.6) in the column 720 determines the weight given to the tag "U2" in the tag-target pair, and the weight is determined to be associated with other factors, such as the tag. The target time, the user confidence assessment (0.7) in the column 725, and then a weight of 0.6, and the correlation score of the tag-target pair is equal to the column correlation (95) The score is multiplied by the weight (0.6) to determine the final correlation score, 57, and in a similar method, the relevance score for the label-target pair located in the column 750 is determined to be 70. * 0.9, or 63, therefore, if a user performs a search request containing the item "U2", the target "Rock Band Home Site", corresponding to the target in the column 750, will be in the reply The list of results (organized) is sorted at a higher position than the target "U2" in the column 740 to indicate that it is more relevant to the user's search.

It can be understood that the file index 700 is only used as an example, different combinations of inputs, different ranges of relevance scores, different algorithms for determining relevance scores (just to list some different architectures), Can be used.

It will be apparent to those skilled in the art that the various modifications of the embodiments can be made without departing from the spirit and scope of the invention as defined by the appended claims.

200. . . flow chart

300. . . member

400. . . flow chart

500. . . flow chart

600. . . Hardware component

Figure 1 is a schematic diagram showing a graphical user interface displaying a list of search results derived at least in part by analyzing tags in accordance with an embodiment of the present invention; Figure 2 is an illustration of According to an embodiment of the present invention, an operational flowchart of an Internet search application capable of applying a label to a file sorting program; FIG. 3 is an illustration of an internet according to an embodiment of the present invention. Schematic diagram of the components of the web search application; FIG. 4 is a flow chart illustrating a procedure for preparing and analyzing tag data according to an embodiment of the present invention; FIG. 5 is an illustration A flowchart of a step of calculating a result using tag data according to an embodiment of the present invention; FIG. 6 is a block diagram illustrating hardware of a component of an Internet search application according to an embodiment of the present invention Figure 7 and Figure 7 show a file index in accordance with an embodiment of the present invention.

Claims (48)

A method of determining the relevance of a plurality of targets to a search request, comprising the steps of: causing one or more of the plurality of targets to be associated with a tag as a result of user input, thereby establishing a tag-target pair Each of the tags includes text; receiving the search request at a search engine; using the association of the tag with the target to calculate a relevance score for each of the plurality of targets for the search request; and responding to the The search request presents a list of results to a user, the list of results including one or more of the plurality of targets, wherein the order of the plurality of targets in the list of results is affected by the relevance score. The method of claim 1, wherein each of the tags from the plurality of tags comprises one or more items, and the method further comprises causing each item from the one or more items to be The target generates an association whereby one or more corresponding project-target pairs are defined. The method of claim 2, further comprising determining a project score for each project-target pair to indicate a degree of relevance between the project and the goal. The method of claim 3, wherein calculating a relevance score for a tag-target pair comprises combining the item scores for the item-target pairs for each item in the tag . The method of claim 4, wherein combining the scores of the items comprises summing the scores of the items. The method of claim 4, wherein combining the item scores comprises weighting each item score by a weight, and summing the weighted item scores. The method of claim 2, wherein the relevance score for a tag-target pair is determined from: the number of occurrences of an item in the tag associated with the target, and the target The number of associated tags, the number of times the tag has been associated with multiple targets, or any combination of these. The method of claim 2, wherein a relevance score for a tag-target pair is determined from: a number of tag-target pairs including one of the tags, including one of the targets Associated tags - the number of target pairs or both. The method of claim 1, wherein the association of the tag with one or more of the plurality of targets is performed by a first user, wherein the tag includes one or more items; The search request includes the one or more items in the tag and is executed by a second user; the result list is presented to the second user; and the relevance score is by the first user or the first A degree of confidence in the confidence of the group to which the user belongs. The method of claim 9, wherein a relevance score for a target and the search request is determined from: items appearing in the target or associated with the search request for the target Correlation score. The method of claim 9, wherein a relevance score from the tag-target pair of the plurality of tag-target pairs is determined from: the first user has made the plurality of goals Any target generates the number of associated tags, the first and the number of targets that the second user has associated with the tag, or both. The method of claim 9, wherein the degree of confidence is a confidence assessment of the first user. The method of claim 12, wherein the confidence assessment is determined from: an assessment of the user of the selection that has associated the target, a search for the first and the second user A measure of similarity between activities, a link metric between the first and the second messenger, or any combination of these. The method of claim 9, wherein the degree of confidence is based on how much a user's tagging behavior conforms to other users' tagging behavior. The method of claim 9, further comprising marking at least one of the plurality of objects in the result list with a graphical component. The method of claim 1, wherein the associating one or more of the plurality of targets with the tag comprises: inputting the tag in an area presented to a user, evaluating the tag , block a link to the target, select the tag, or select the target. The method of claim 1, wherein causing one of the plurality of targets to be associated with the tag comprises: analyzing an input given by a user performing a search. The method of claim 1, wherein the plurality of targets include a hyperlink to a webpage or a hyperlink group pointing to a webpage. The method of claim 1, wherein the plurality of target systems include: pointing to text, images, photos, tags, tag groups, subject areas, concepts, user profiles, answers, audio files, video File, software, or any combination of these. The method of claim 1, wherein a tag crawler associates at least one of the plurality of tags with at least one of the plurality of targets. The method of claim 1, wherein the step of presenting the result list to the user comprises displaying the tag associated with one or more of the plurality of targets in the result list to the user . A method of establishing a system for presenting a goal to a user, comprising the steps of: storing a plurality of tags associated with a plurality of targets in a tag database as a result of user input; A score database stores a relevance score between the plurality of tags and the plurality of targets, wherein the relevance scores are used to organize a plurality of files in an organized result list. The method of claim 22, wherein storing a plurality of tags in a tag database comprises storing items from the plurality of tags, and wherein storing the relevance scores comprises rating the items and A correlation score for a correlation between these goals. The method of claim 23, further comprising storing a plurality of indexes in the index database, each index corresponding to an item from the plurality of items, a corresponding target from the plurality of targets, And a corresponding relevance score between the project and the goal. The method of claim 24, wherein each corresponding relevance score between an item and a target is affected by a degree of confidence of a person or a group to which the person belongs, the person performing the The association between the tag and the target. The method of claim 24, wherein a relevance score between a project and a goal is determined by a statistical taxonomy or a sorting regression algorithm. The method of claim 26, wherein the statistical classification method or the sorting regression algorithm is a logistic regression, a support vector machine, a classification or regression tree, and a lifting tree. Any of the boosted tree ensembles. The method of claim 22, further comprising: presenting a list of results to a user in response to a search request including a target; and the project is included in the result list by the user A target in the relationship is generated; and a relevance score between the item and the target is determined. The method of claim 28, wherein a relevance score between a target from the plurality of targets and a tag associated therewith is determined from one or more of the following: the tag has been With the item The number of times the association is generated, the total number of labels associated with the target, the number of times the label has been associated with any of the multiple targets, the number of labels that have been associated with all of the multiple targets, the label The time associated with the target, the number of times the target has been bookmarked, the value and quantity applied to the target, or any combination of these. A method for organizing a plurality of objects for display in a result list, comprising the steps of: correlating items in a search request with tags associated with a plurality of targets as a result of user input; And replying to a list of results comprising the plurality of goals organized based on the associations, each association corresponding to a correlation metric. The method of claim 30, further comprising: performing the search request, thereby generating a first target list; and organizing the plurality of targets based on the associations. The method of claim 31, further comprising associating the tag with the target. The method of claim 30, further comprising applying a statistical classification, or sorting a regression algorithm to the plurality of targets to determine a measure of correlation between the plurality of targets and the item of the search request. The method of claim 33, wherein the statistical classification or the sorting regression algorithm is a logistic regression, a support vector machine, a classification or regression tree, and a lifting tree as a whole ( Boosted tree ensembles). A system that responds to a search request and responds to a list of search results The system includes: a tag database for storing a tag associated with the target, wherein one or more targets are associated with the tag as a result of the user input, and each tag includes text; a tag analyzer, coupled Connecting to the tag database, wherein the tag analyzer is configured to use the association of the tag with the target to calculate a relevance score for each of the targets for the search request; and a search engine for Receiving the search request and for presenting a list of results, the search result including a reference to one or more of the targets, wherein the order of the plurality of targets in the result list is affected by the relevance score . The system of claim 35, further comprising a target index for storing a relevance score between the tag and the target associated with the target, between the bookmark and the target, or both. The system of claim 35, wherein the relevance score is determined by summing the items used to form a label and the weighted relevance score of the target. The system of claim 35, wherein the relevance score between a search request including a project and a goal is determined by: the number of tags included in the search request, the search request The number of times a tag is included in the tag database, the number of tags associated with the target, the number of items in the tag and the items in the search request, the target has been bookmarked The number of times, the number of times the target has been evaluated, or any combination of these. The system of claim 35, wherein the relevance score of a tag and a target is based on: a position of the tag within the target, a frequency of the tag within the target, the tag is at the A density within the target, or any combination of these. The system of claim 35, wherein the weighting of the relevance score for a tag and a target is based on a degree of confidence assigned to a user who has the tag and the target Generate an association. The system of claim 35, wherein the relevance score is a statistical classification, or a sorting regression algorithm, a clustering analysis algorithm, or a type analysis algorithm ( Morphological analysis algorithm). The system of claim 41, wherein the statistical classification or sorting regression algorithm comprises a logistic regression, a support vector machine, a classification or regression tree, and a lifting tree as a whole ( Boosted tree ensembles). The system of claim 36, further comprising a search engine coupled to the target index, wherein the search engine is programmed to receive a search request, the search request including an item corresponding to the label, And replying to the list of results of the organization based on the relevance score of the tag-target pair. The system of claim 43, further comprising a user database coupled to the search engine, the user database including information related to the search request. The system of claim 44, wherein the information related to the search request includes a link tracked by the user, a tag associated with the target, a target blocked by the user, a bookmark, or the like. Any combination. The system of claim 35, wherein the target comprises a hyperlink to a webpage or a hyperlink to a webpage group. The system of claim 35, wherein the objectives include pointing to text, images, photos, tags, tag groups, subject areas, concepts, user profiles, answers, audio files, Video files, software, or hyperlinks to any combination of these. The system of claim 35, further comprising a means for associating a tag with a target.