MXPA00008603A - Identifying the items most relevant to a current query based on items selected in connection with similar queries - Google Patents

Abstract

The present invention provides a software facility for identifiying the items most relevant to a current query based on items selected in connection with similar queries. In preferred embodiments of the invention, the facility receives a query specifiying one or more query terms. In response, the facility generates a query result identifying a pluralityof items that satisfy the query. The facility then produces a ranking value for at least a portion of the items identified in the query result by combining the relative frequencies with which users selected that item from the query results generated from queries specifying each of the terms specified by the query. The facility identifies as most relevant those items having the highest ranking values.

Description

IDENTIFICATION OF THE MOST RELEVANT ITEMS FOR A CURRENT CONSULTATION BASED ON ARTICLES SELECTED WITH RESPECT TO SIMILAR CONSULTATIONS TECHNICAL FIELD The present invention is directed to the field of query processing.

BACKGROUND OF THE INVENTION Many global websites allow users to search to identify a small number of interesting articles among a domain of much larger articles. As an example, several web index sites allow users to search for particular websites among most known websites. Similarly, many online merchants, such as book sellers, allow users to search for particular products among all products that can be compared by a merchant. In many cases, users perform searches in order to finally find an individual article within a complete domain of articles. In order to perform a search, a user makes a query containing one or more query articles. It also explicitly or implicitly identifies an article domain to search. For example, a user can submit a query to an online book seller containing terms that the user believes are words in the title of a book. A query server program processes the query to identify within the domain, articles that match the terms of the query. The items identified by the query server program are collectively known as a result of queries. In the example, the query result is a list of books whose titles contain some or all of the terms of the query. The query result is typically displayed to the user as a list of items. This list can be ordered in several ways, for example, the list can be ordered alphabetically or numerically based on a property of each item, such as the title, author, or production date of each book. As another example, the list can be ordered based on the degree to which each identified item matches the terms of the query. When the domain for a query contains a large number of articles, it is common for query results to contain 10 or hundreds of articles. When the user is performing the search in order to find a single item, the application of conventional aspects to order the query result usually fails to place the desired article or articles near the top of the query result, so that the user must read through many other articles in the query result before reaching the searched article. In view of this disadvantage of conventional aspects for ordering query results, a new, more effective technique for automatically ordering query results according to the behavior of the collective and individual user could have an important utility. In addition, it is very common for them to specify queries that are not satisfied by any article. This can happen, for example, when a user makes a detailed query that is very close, or when a user writes badly or does not remember well an article in the query. In such cases, conventional techniques, which only present articles that satisfy the query, do not present any article to the user. When no article is presented to a user in response to the issuance of a query, the user may become frustrated with the search engine, and may still discontinue its use. Therefore, a technique for displaying articles that relate to at least some of the terms to a query even if no article completely matches the query, could have an important utility. In order to satisfy this need, some search engines adopt a strategy to effectively and automatically review the query until a non-empty result group is produced. For example, a search engine can progressively eliminate conjunctive terms, ie, ANDed, from a multiple-term query until the result group produced for that query contains the articles. This strategy has the disadvantage that the important information to select the correct articles can be lost when the articles of consultation are arbitrarily eliminated. As a result, the first group of non-empty results may be too large, and may contain a large percentage of items that are not relevant to the original query as a whole. For this reason, a more effective technique for displaying articles in relation to at least some of the terms in a query, even if no article completely matches the query, could have an important utility.

COMPENDIUM OF THE INVENTION The present invention provides a software installation ("the installation") to identify the most pertinent articles for a current query based on selected articles with respect to similar queries. The installation preferably generates classification values for articles indicating their level of applicability for the current query, which satisfies one or more query terms. The installation generates a classification value for an article by combining classification notes, produced by a classification function, each corresponding to the level of applicability of the article for queries containing one of the classification values. The classification function preferably retrieves a classification note for the combination of an article and a term from a classification table generated by the installation. The notes in the classification table preferably reflect, for a particular article and term, how, in general, users have selected the article when the article has been identified in the query results produced for queries that contain particular terms. In different modalities, the installation uses the classification notes either to generate a classification value for each article in a query result, or to generate classification values for a smaller number of articles in order to select some items that have the higher classification values. To generate a classification value for a particular article in a query result, the installation combines the classification notes that correspond to that article and the query terms. In modalities where the objective is to generate classification values for each article in the query result, the installation preferably connects the articles in the query results and, for each article, combines all the classification notes corresponding to that article and any of the articles in the query. On the other hand, in modalities where the objective is to select some articles in the query result having the highest classification values, the installation preferably connects the articles in the query, and, for each article, identifies some of the classification notes higher for that term and any article. The installation then combines the notes identified for each item to generate classification values for a relatively small number of items, which may include items not identified in the query result. In fact, these embodiments of the invention are capable of generating classification values for and displaying articles even in cases where the query result is empty, that is, when no article completely satisfies the query. Once the installation has generated classification values for at least some items, the facility preferably orders the items of the result in descending order of sort value. The installation can also use the classification values to form a subset of items in the query result to a smaller number of items. By sorting and / or subsetting the articles in the query result in this way according to the behavior of the collective and individual user instead of according to the attributes of the articles, the installation substantially increases the probability that the user will quickly find within the query result the particular article or articles that it is looking for. For example, while a query result for a query containing the terms "human" and "dynamic" may contain a book regarding human dynamics and a book regarding the effects on humans of particle dynamics, the selections by users of the results of previous queries produced for queries containing the term "human" show that these users select the human dynamics book much more frequently than selecting the particle dynamics book. Therefore, the installation classifies the book of human dynamics greater than the book of particle dynamics, allowing users who are more interested in the book of human dynamics to select it more quickly. This benefit of the installation is especially useful along with the larger, heterogeneous query results that are typically generated for queries of individual terms, which are commonly performed by users. Several embodiments of the invention base the classification notes on different types of selection actions performed by users on articles identified in query results. These include if the user displays additional information regarding an article, how much time the user spends on seeing the additional information regarding the article, how many hyperlinks the user follows within the additional information regarding the article, if the user adds the article to your shopping basket, and if the user finally bought the item. The embodiments of the invention also consider selection actions that are not related to query results, such as writing an article identifier of the article instead of selecting the article from a query result. Additional embodiments of the invention are incorporated into the classification process information with respect to the user making the query by maintaining and applying separate classification notes for users in different demographic groups, such as those of the same sex, age, income, or geographic category. . Certain modalities also incorporate behavioral information regarding specific users. In addition, classification notes can be produced through a classification function that combines different types of information that reflect collective and individual user references. Some embodiments of the invention use specialized strategies to be incorporated into the classification note information with respect to queries made at different time points.

BRIEF DESCRIPTION OF THE DRAWINGS Figure 1 is a high level block diagram showing the computer system where the installation is preferably executed. Figure 2 is a flow diagram showing the steps preferably performed by the installation in order to generate a new classification table. Figures 3 and 4 are table diagrams showing an increase of an article classification table according to step 206 (Figure 2).

Figure 5 is a table diagram showing the generation of classification tables for composite time periods from classification tables for constituent time periods. Figure 6 is a table diagram showing a classification table for a composite period. Figure 7 is a flow diagram showing the steps preferably performed by the installation in order to identify the user selections within a web server registry. Figure 8 is a flow diagram showing the steps preferably performed by the facility to order a query result using a classification table generating a classification value for each article in the query result. Figure 9 is a flow chart showing the steps preferably analyzed by the facility to select some items in a query result having the highest ranking values using a classification table.

DETAILED DESCRIPTION OF THE INVENTION The present invention provides a software installation ("the installation") to identify the most pertinent articles for a current query based on selected articles with respect to similar queries. The installation preferably generates classification values for articles indicating their level of applicability for the current query, which identifies one or more query terms. The installation generates a classification value for an article by combining classification notes, produced by a classification function, that each corresponds to the level of applicability of the article for the queries containing one of the classification values. The classification function preferably retrieves a classification note for the combination of an article and a term from a classification table generated by the installation. The notes in the classification table preferably reflect, for a particular article and term as, in general, users have selected the article when the article has been identified in the results of queries produced for queries containing the term. In different modalities, the installation uses the classification notes either to generate a classification value for each article in a query result, or to generate classification values for a smaller number of articles in order to select some articles having the Higher classification values. To generate a classification value for a particular article in a query result, the installation combines the classification notes that correspond to that article, and the query terms. In modalities where the objective is to generate classification values for each article in the query result, the facility preferably connects the articles in the query results and, for each article, combines all the classification notes corresponding to that article and any of the terms in the query. On the other hand, in modalities where the objective is to select some articles in the query result having the largest classification values, the installation preferably connects the terms in the query, and, for each article, identifies some higher ranking notes for that term and any article. The installation then combines the notes identified for each item to generate classification values for a relatively small number of items, which may include items not identified in the query result. In fact, these embodiments of the invention are capable of generating classification values for and displaying articles even in cases where the query result is empty, that is, when no item complements satisfies the query.

Once the installation has generated classification values for at least some items, the facility preferably orders the items of the query result in descending order of classification value. The installation can also use the classification values to form a subset of the items in the query result to a number of smaller items. Through the arrangement and / or formation of subgroups of the articles in the query result in this way, according to a collective and individual user behavior, instead of according to attributes of the articles, the installation substantially increases the probability that the user will quickly find within the query result the article or particular articles that he is looking for. For example, while a query result for a query containing the terms "human" and "dynamic" query may contain a book regarding human dynamics and a book regarding humans of particle dynamics, selections by Part of the users of the previous query results produced for queries containing the term "human" show that these users select the book of human dynamics much more frequently than those who select the book of particle dynamics. Therefore, the installation classifies the book of human dynamics greater than the book of particle dynamics, allowing users, most of whom are more interested in the book of human dynamics to select it more easily. This benefit of the installation is especially useful along with the large, heterogeneous query results that are typically generated for individual term queries, which are commonly made by users. Several embodiments of the invention are based on classification notes in different types of selection actions performed by users on items identified in query results. These include whether the user displays additional information regarding an article, how much time the user spends on seeing the additional information on the article, how many hyperlinks the user follows within the additional information regarding the article, if the user adds the article to your shopping basket, and if the user finally buys the item. The embodiment of the invention also considers selection actions unrelated to the query results, such as writing an article's article identifier instead of selecting the article from a query result. Additional embodiments of the invention are incorporated into the classification process information with respect to the user making the query by maintaining and applying separate classification notes for users in different demographic groups, such as those of the same sex, age, income, or geographic category. . Certain modalities also incorporate behavioral information regarding specific users. In addition, classification notes can be produced by a classification function that combines different types of information that reflects collective and individual user preferences. Some embodiments of the invention use specialized strategies to incorporate classification information into information regarding queries made in different time frames. Figure 1 is a high level block diagram showing the computer system on which the installation is preferably executed. As shown in Figure 1, the computer system 100 comprises a central processing unit (CPU) 110, input / output devices 120, and a computer memory (memory) 130. Among the input / output devices it is found a storage device 121, such as a hard disk drive; a computer-readable medium unit 122, which can be used to install software products, including installation, which are provided on a computer-readable medium, such as a CD-ROM; and a network connection 123 for connecting the computer system 100 to other computer systems (not shown). The memory 130 preferably contains a query server 131 for generating query query results, a query result classification facility 132 for automatically classifying the articles into a query result according to collective user preferences, and classification tables of articles 133 used by the installation. Although the installation is preferably implemented in a computer system configured as described above, those skilled in the art will recognize that it can also be implemented in computer systems having different configurations. The installation preferably generates a new classification table periodically, and when a query result is received, it uses the classification table finally generated to classify the articles in the query result. Figure 2 is a flow diagram showing the steps preferably performed by the installation in order to generate a new classification table. In step 201, the installation initiates a classification table to maintain the entries, each indicating the classification note for a particular combination of a query term and an article identifier. The leaderboard preferably has no entries where it is initialized. In step 202, the installation identifies all the query result item selections made by users during the period in which the classification table is being generated. The classification table can be generated for queries that occur within a period of time, such as 1 day, 1 week or 1 month. This group of queries is referred to as a "classification group" of queries. The installation also identifies the terms of the queries that produce these query results in step 202. The implementation of step 202 is discussed in more detail below in relation to FIG. 7. In steps 204-208, the installation connects each Article selection of a query result that was made by a user during the time period. In step 204, the facility identifies the terms used in the query that produced the query result where the article selection was presented. In steps 205-207, the installation connects each term in the query. In step 206, the installation increases the classification notes in the classification table corresponding to the current article and term. When an entry does not yet exist in the classification table for the term and article, the installation adds a new entry to the classification table for the term and the article. Increasing the classification notes preferably involves adding an increment value, such as 1, to the existing classification notes for the term and the article. In step 207, if the additional terms remain processed, the installation is returned to step 205 to process the next term in the query, and also the installation continues in step 208. In step 208, if the additional item selections remain in process, then the installation returns to step 203 to process the next article selection, also these steps conclude. Figures 3 and 4 are table diagrams showing an increase of an article classification table according to step 206 (Figure 2). Figure 3 shows the status of the article classification table before its increase. It can be seen that table 300 contains a number of entries, including entries 301-306. Each entry contains the classification note for a particular combination of a query term and an article identifier. For example, entry 302 identifies the note or rating "22" for the term "dynamic", the item identifier "1883823064". You can, see by examining entries 301-303 that, in query results produced from queries including the term "dynamic", the article that has the item identifier "1883823064" has been selected by users more frequently than the article that has the item identifier "9676530409" and much more frequently than the article that has the item identifier "0801062272".

In additional modalities, the installation uses several other data structures to store classification notes, such as sparse dispositions. By increasing the article classification table 300, the installation identifies the selection of the article that has the item identifier "1883823064" from a query result produced by a query specifying the query terms "human" and "dynamic". Figure 4 shows the status of the article classification table after the article classification table is augmented by the installation to reflect this selection. It can be seen by comparing entry 405 in the classification table of article 400 with entry 305 in the classification table of article 300 that the installation has increased the notes for this entry from "45" to "46". Similarly, the installation has increased the classification notes for this article identifier, the term "dynamic" from "22" to "23". The installation increases the classification table in a similar way for the other selections for the query results that are identified during the time period. Instead of generating a new classification table from the trace using the steps shown in Figure 2, each time a new selection information becomes available, the installation preferably generates and maintains separate classification tables for different constituent time periods , of a relatively short length, such as 1 day. Each time a classification table is generated for a new constituent time period, the facility preferably combines this new classification table with existing classification tables for previous constituent time periods to form a classification table over a more composite time period. long. Figure 5 is a table diagram showing the generation of classification tables for time periods composed of classification tables during constituent time periods. You can see in Figure 5 that the classification tables 501-506, each corresponds to a single day between February 8, 1998 and February 13, 1998. Each time a new constituent period is completed, the installation generates a new classification table reflecting user selections made during that constituent period. For example, at the end of February 12, 1998, the installation generates the 505 classification table, which reflects all user selections that occurred on February 12, 1998. After the installation generates a new classification table during a completed constituent period, the facility also generates a new classification table during a compound period ending with that constituent period. For example, after the generation of the 505 classification table for the constituent period of February 12, 1998, the facility generates the classification table 515 for the period comprised from February 8, 1998 to February 12, 1998. The installation preferably generates said classification table during a compound period by combining the entries in the classification table for the constituent periods that make up the composite period, and combining the notes of corresponding entries, for example, by adding them. In a preferred embodiment, the notes and classification tables for more recent constituent periods are loaded more heavily than those in the classification tables for less recent constituent periods. When classifying query results, the classification table for the most recent composite period is preferably used. That is, until the classification table 516 can be generated, the installation preferably uses the classification table 515 to classify query results. After the classification table 516 is generated, the installation preferably uses the classification table 516 to classify query results. The durations of both constituent periods and composite periods are preferably configurable. Figure 6 is a table diagram showing a classification table for a composite period. By comparing the classification table of article 600 shown in Figure 6 with the classification table of article 400 shown in Figure 4, it can be seen that the contents of classification table 600 constitute the combination of the contents of the table of 400 classification with several other classification tables for constituent periods. For example, the classification note for entry 602 is "116", or approximately 5 times the classification for the corresponding entry 402. In addition, although the classification table 400 does not contain an entry for the term "dynamic" and the identifier of item "1887650024", entry 607 has been added to table 600 for this combination of term and item identifier, as a corresponding entry occurs in the classification table for one of the other constituent periods within the composite period. The process used by the facility to identify user selections depends both on the type of selection action used by the installation and the manner in which the data in relation to said selection actions are stored. A preferred embodiment uses as its selection action requests to display more information regarding articles identified in query results. In this mode, the installation extracts this registration information generated by a web server that generates query results for a user using a web client, and allows the user to select an article with the web client in order to display additional information with with respect to it. A web server generally maintains a record that details all HTTP requests that have been received from web clients and answered. Said registry is generally made of entries, each containing information with respect to a request other than HTTP. These records are usually organized chronologically. Registry entry 1, below, is a sample registration entry illustrating an HTTP request made by a web client in favor of the user making a query. 1. Friday, 13-Feb-98 16:59:27 2. User ID = 82707238671 3. HTTP_REFERER = http: // www. amazon.com/book_query_page 4. PATH_INFO = / libro_consulta 5. author = "Seagal" 6. title = "Human Dynamics" Registration entry 1 It can be seen that the occurrence of the key "book_consulta" in the line of "PATH_INFO" 4 of the record entry 1 that this record entry corresponds to an introduction of a query made by a user. In addition, it can be seen in the lines of terms 5 and 6 that the query includes the terms "Seagal", "Human" and "Dynamic". In line 2, the entry also contains a user identifier that corresponds to the user's identity and, in some modalities, also to this particular interaction with the web server. In response to receiving the HTTP request documented in record entry 1, the query server generates a query result for the query and returns it to the web client by making the query. Subsequently, the user selects an item identified in the query result, and the web client makes another HTTP request to display detailed information regarding the selected article. Registry entry 2, which occurs at a point after record entry 1, in the record, describes this second HTTP request. 1. Friday, 13-Feb-98 17:02:39 2. User ID 82707238671 3. HTTP_REFERER = http: // www. amazon.com/book_query_page 4. PATH_INFO = / ISBN = 1882823064 Registration Entry 2 By comparing the user identifier on line 2 of record entry 2 with the user identifier on line 2 of record entry 1, it can be seen that these record entries correspond to the same user and time frame. In the line "PATH_INFO" 4 of record entry 2, it can be seen that the user has selected an article that has an article identifier ("ISBN") "1883823064". In addition, it can also be seen from the occurrence of the key "libro_consulta" on the line "HTTP_REFERER" 3 that the selection of this article was from a query result. When the information regarding user selections is stored in web server registers, such as those discussed above, the installation preferably identifies user selections through these registers. Such a crossover may occur either in an intermittent processing mode after a record for a specific period has been fully generated, or in a real-time processing mode, such that those record entries are processed as rapidly as they are generated. . Figure 7 is a flow chart showing the steps preferably performed by the installation in order to identify user selections within a web server registry. In step 701, the installation places a first flag at the top, or beginning, of the record. The installation then repeats steps 702-708 until the first prompt reaches the end of the record. In step 703, the installation proceeds forward with the first flag to the next item selection event. In terms of the record entry shown above, step 703 involves moving forward through the registry entries until one is found that contains on its "HTTP_REFERER" line a key denoting a search entry, such as "book_consultation" . In step 704, the extra installation of this article selection event, the identity of the article that was selected and the session identifier that identifies the user of the selected article. In terms of the previous registry entries, this involves reading the 10-digit number followed by the "ISBN-" strip on the "PATH NFO" line of the registry entry, and reading the user identifier from the line of "User Identifier" of the registry entry. In this way, in record entry 2, the installation extracts the item identifier "1883823064" and the session ID "82707238761". In step 705, the installation synchronizes the position of the second flag with the position of the first flag. That is, the installation causes the second flag to signal to the same registry entry as the first flag. In step 706, the installation traverses backward with the second pointer to a query event having a matching user identifier. In terms of the previous registry entries, the installation crosses back to the registry entry having the key "libro_consulta" in its line "PATH_INFO", and having a matching user identifier in its "User Identifier" line. In step 707, the installation extracts from the query event where the second flag points out the terms of the query. In terms of the previous query record entries, the installation extracts the words in quotes from the query record entry where the second signaling signal occurs, in the lines after the "PATH_INFO" line. In this way, in the registry entry 1, the installation extracts the terms "Seagal", "Human" and "Dynamic". In step 708, if the first flag has not reached the end of the record, then the installation returns to step 702 to continue processing the record, and also these steps conclude. When other selection actions are used by the installation, the extraction of information with respect to the information regarding selection of the web server registry may be a little more involved. For example, when the installation uses the purchase of the item as the selection action, instead of identifying a registration entry describing a request by the user for more information regarding an item, such as the record entry 1, the installation more well identifies a registry entry that describes a request to buy items in a shopping basket. The facility then crosses back into the registry, using the entries that describe requests to add items and remove items from the shopping basket to determine which items were in the shopping basket at the time of the purchase request. The installation then continues to cross back into the registry to identify the registry entry that describes the query, such as registry entry 2, and to extract the search terms. Instead of relying solely on a web server registry, where the article purchase is the selection action that is used by the installation, the installation alternately uses a separate database from the web server registry to determine which items are purchased. in each purchase transaction. This information from the database is then matched to the record entry that contains the query terms for the query for which the item is selected for purchase. This hybrid aspect, using the web server registry and a separate database, can be used for any of the different types of selection actions. In addition, when a database separated from the web server registry contains all the information necessary to increase the classification table, the installation can use the database exclusively, and avoid crossing the web server registry. The installation uses classification tables that have been generated to develop classification values for articles in new query results. Figure 8 is a flow chart showing the steps preferably performed by the facility to order a query result using a classification table and generating a ranking value for each article in the query result. In steps 801-807, the installation connects each item identified in the query result. In step 802, the installation initializes a classification value for a current article In steps 803-805, the installation connects each term that occurs in the query. In step 804, the installation determines the classification notes contained by the most recently generated classification table for the current term and articles. In step 805, if any of the terms of the query remain in process, then the installation continues to step 803, in addition the installation continues in step 806. In step 806, the installation combines the notes for the current article to generate a classification value for the article. As an example, referring to Figure 6, when processing data that has an item identifier "1883823064", the facility combines the value "116" extracted from entry 602 for this article and the term "dynamic" and the value " 211"extracted from entry 605 for this article and the term" human ". Step 806 preferably involves adding these values. These values can be combined in other ways, however. In particular, the values can be adjusted to more directly reflect the number of query terms that match the article. So that the articles that coincide with more terms of consultation than with others are favored in the classification. In step 807, if any article remains in process, the installation returns to step 801 to process the next article, also the installation continues in step 808. In step 808, the installation displays the items identified in the query result of according to the classification values generated by the articles in step 806. Step 808 preferably involves classifying the articles in the query result in descending order of their classification values, and / or forming subgroups of the articles in the query result. to include only those items above a threshold rating value, or only a predetermined number of items having the highest rating values. Then, from step 808, these steps conclude. Figure 9 is a flow chart illustrating the steps preferably performed by the facility to select some items in a query result having the highest ranking values using a classification table. In steps 901-903, the installation connects each term in the query. In step 902, the facility identifies between the entries in the table the current term and those entries that have the three highest rating values. For example, with reference to Figure 6, if the only entries in the article 600 classification table for the term "dynamic" are entries 601, 602, 603, and 607, the installation could identify entries 601, 602, and 603, which are the entries for the term "dynamic" that has the three highest ranking values. In additional preferred embodiments, a small number of table entries is used instead of three. In step 903, if the additional terms remain in the query that is being processed, then the installation returns to step 901 to process the next term in the query, also in the continuous installation in step 904. In steps 904-906 , the installation connects each unique article among the identified entries. In step 905, the installation combines all the values for the article between the identified entries. In step 906, if additional unique items remain among the identified entries that are being processed, then the installation returns to step 904 to process the next single item, also the continuous installation in step 907. As an example, yes, in the article classification table 600, the installation selected entries 601, 602, and 603 for the term "dynamic", and the selected entries 604, 605 and 606 for the term "human", then the installation can combine the values "116" and "211" for the article that has the item identifier "1883823064", and can use the following individual values for the article identifiers remaining: "77" for the item that has the item ID "0814403484", "45" for the item that has the item ID "9676530409", "12" for the item that has the item ID "6303702473", and "4" for the article that has the item identifier "0801062272". In step 907, the installation selects prominent display items that have the top three values combined. In additional modalities, the installation selects a small number of items that have higher combined values that are different from 3. In the example discussed above, the facility could select the prominent display of items that have item identifiers "1883823064", "0814403484"and" 9676530409". Since the installation in step 907 selects items without considering their presence in the query result, the installation can select article that is not in the query result. This aspect of this mode is particularly advantageous in situations where a full query result is not available when the installation is invoked. As in the case, for example, where the query server only provides a portion of the items that satisfy the query at that time. This aspect of the invention is also advantageous since, by selecting articles without considering their presence in the query result, the installation is able to select and display the user articles in relation to the query, even when the query result is empty, it is say, when no item completely satisfies the query. After step 907, these steps conclude. Although the present invention has been shown and described with reference to preferred embodiments, it will be understood by those skilled in the art that various changes or modifications in form and detail may be made without departing from the scope of the invention. For example, the installation can be used to classify query results of all types. The installation can use several formulas to determine, in the case of each article selection, the amount by which to increase the classification values with respect to the selection. In addition, the installation can use several formulas to combine classification values into a classification value for an article. The installation can also use a variety of different types of selection actions to increase the classification table, and can increase the classification table for more than one type of selection action at a time. In addition, the installation may increase the classification table to reflect selections by users other than human users, such as software agents and other types of artificial users.

Claims (12)

1. - A method for being used in a computer system (100) of classification articles in a current search result, the method comprises: for each of the multitude of search terms, compile (201 to 208) data (300) indicating the degree to which the users have selected each of the multitude of articles (301 to 306) when they return in previous search results produced from previous queries containing the search term; receive a current query and a current search result, the current query received containing one or more terms from the multitude of terms, the current search result received identifying a plurality of items among the multitude of items that satisfy the current query received; and classifying (801 to 808), using the compiled data (300), at least a portion of the items identified in the current search result received in accordance with the degree to which the users have selected each of the plurality of items identified in the current search result received when it returns to previous search results produced from previous queries containing one or more of the search terms contained in the current query received.

2. The method according to claim 1, wherein at least a portion of the users is identified with one of a plurality of demographic groups; the compilation step (201 to 208) compiles, for each demographic group of the plurality of demographic groups, data indicating the degree to which the users identified with the demographic group have selected each of the multitude of articles when they return in search results previous ones produced of previous consultations containing the searched term; the current query received is presented in favor of a distinguished user identified with a distinguished demographic group; and the classification step (801 to 808) uses the compiled data (300) to classify the items identified in the previous search result received according to the degree to which the users identified with the distinguished demographic group have selected each of the plurality of articles identified in the current search result received when they return in previous search results produced from previous queries containing the search term contained in the received query.

3. The method according to claim 1 or 2, further comprising imposing the items identified in the current search result received in an order in which the item classification values monotonically is reduced.

4. The method according to one of claims 1 to 3, further comprising creating (901 to 907) an appropriate subgroup of the articles using the classification values of the articles identified in the received current search result.

5. - The method according to claim 4, wherein the step of creating (901 to 907) creates a subset of the articles identified in the received current search result that contains all the items whose classification values exceed a minimum classification value predetermined.

6. The method according to one of claims 1 to 5, which further comprises increasing (701 to 708) classifications for selections made to display additional information with respect to the articles.

7. The method according to one of claims 1 to 6, which further comprises increasing (701 to 708) the classifications for the selections made to purchase items.

8. The method according to one of claims 1 to 7, which further comprises increasing (701 to 708) the classifications for selections made to add items to a tentative shopping list.

9. The method according to one of claims 1 to 8, further comprising, increase (701 to 708) the classifications for the selections of detailed information portions displayed with respect to items.

10. The method according to one of claims 1 to 9, further comprising: increasing (701 to 708) classifications for units of time for which the user displays detailed information regarding articles.

11. A computer system (100) adapted to perform the method of one of claims 1 to 10. 12.- A computer-readable medium whose contents cause a computer system (100) to perform the method of one of the claims 1 to 10.