Claims:1. A computer-implemented method function at a server system having one or more processors and memory, the method comprising:
a. receiving a search query from a user;
b. identifying a list of search results associated with the search query
c. identifying a user profile associated with the user,
d. identifying a set of candidate search results in a search history of the user
e. determining a popularity metric for each of the candidate search results; and
f. selecting a subset of the candidate search results
g. identifying in the list of search results, one or more search results that are associated with at least one of the user-preferred search results;
h. reordering the list of search results by moving each of the identified search results from its initial position by an offset 2.
2. The method of claim 1, wherein the offset is a variable that is a function of popularity metric associated with each of the identified search results.
3. The method of claim 1, wherein each of the candidate search results has been selected by the user for at least the predefined minimum number of times
4. The method of claim 1, wherein the popularity metric of a candidate search result is a function of one or more parameters
5. The method of claim 1, wherein the popularity metric of a candidate search result including at least one parameter which is a time span period from the user's most remote selection of the candidate search result to the user's most recent selection of the candidate search result.
, Description:Technical Field of the Invention
The present invention relates generally to the field of search engines in computer network systems, to systems and methods of personalizing search results using a user's search history.
Background of the Invention
Search engines are powerful tools for locating and retrieving documents from the Internet (or an intranet). Traditionally, different users who submit the same search query to a search engine receive the same set of search results. For example, the search engine generates the same set of search results for the search query “apple” irrespective of whether the search query is from a user who is interested in Apple® computers or from a user who is interested in the fruit malus domestica. Clearly such search results are likely to include some results of little interest to either user.
In view of the, it would be desirable to have a search engine that can customize its search results to highlight information items in the search results that are most likely to be of interest to users who submit the search queries. Further, it would be desirable for such a system to operate without explicit input from a user regarding the user's personal preferences and interests, and for the system to protect the privacy interests of its users.
Object of the Invention
The primary object of present invention is to provides Methods and System to relegate search result based on personal profile.
Summary of the Invention
In some embodiments, an information server identifies a list of search results associated with a search query from a user. Each search result has an initial position in the list. The information server then identifies a user profile associated with the user. In some embodiments, the user profile includes a set of user-preferred search results that is determined at least in part from the user's search history. Among the list of search results, the information server then identifies one or more search results that are associated with at least one of the user-preferred search results. Finally, the information server reorders the list of search results by moving each of the identified search results from its initial position by an offset and then provides the reordered list of search results to the user. Some embodiments may be implemented on either the client side or the server side of a client-server network environment.
Brief Description of the Drawings
FIG. 1 is a process for personalizing search results using a user profile in accordance with some embodiments.
Detailed Description of Invention
FIG. 1 is a flowchart process for personalizing search results using a user profile in accordance with some embodiments. Initially, the search result ranker identifies a list of search results associated with a search query from a user 1. As noted above, the list of search results is produced by the search engine. While processing the search query, each search result in the list has an initial position that is dependent at least in part by the search result's generic ranking score.
Next, the search result ranker identifies a user profile associated with the user (2). In some embodiments, the user may have multiple user profiles, such as a set of user-preferred search results and a set of user-disfavoured search results. These two sets of search results are selected from the user's search history record. In some embodiments, a search result cannot appear in both sets. In some other embodiments, a search result appearing in both sets is treated as a user-preferred search result to avoid removing or demoting a user-preferred search result by accident.
The search result ranker then compares the list of search results with the user-preferred search results and/or the user-disfavoured search results to identify M user-preferred search results and/or M' user-disfavoured search results associated with each set (3). Note that the number M or M' will vary from one set of search results to another.
In some embodiments, a search result is associated with the user-preferred search results if it is a member of the user-preferred search results. In particular, the search result is deemed to be a member of the user-preferred search results if its URL is identical to a URL of one of the user-preferred search results. In some embodiments, a search result is associated with the user-preferred search results if it is associated with a source shared by at least one of the user-preferred search results. For example, different URLs associated with the same website all include the website's hostname, which is deemed to be the source of the URL. In some embodiments, the user profiler only looks at the hostname of a URL when choosing the user-preferred search results from the user's search history record. As a result, any search result whose URL includes one of the user-preferred hostnames is deemed to be associated with the user-preferred search results.
The positions of the M identified user-preferred search results and/or M' user-dis favoured search results (4). For example, the search result ranker moves each of the identified search results from its current position determined by its generic ranking score by an offset, based on a presumption that a user-preferred search result near the top of the list is likely to receive more attention from the user while a user-dis favoured search result near the bottom of the list is likely to receive less attention from the same user.
In some embodiments, the offset is a constant for the M or M' identified search results. For example, the search result ranker moves each of the M identified user-preferred search results a few positions (e.g., one or two positions) upward in the list and moves each of the M' identified user-disfavoured search results a few positions (e.g., one to five positions) downward in the list.
In some embodiments, the offset is a variable that depends at least in part on the popularity metric (or unpopularity metric) associated with each of the M (or M') identified search results. For example, a search result is moved upward/downward by several positions proportional to its popularity metric (or unpopularity metric). A search result with a relatively higher popularity metric is moved more positions upward than another one with a relatively lower popularity metric. In some embodiments, the search result ranker multiplies a search result's generic ranking score by its popularity metric to determine the position offset for the search result. In some embodiments, the popularity metric of a user-preferred search result is a prediction of the search result's long click-through rate, which indicates the likelihood of a user selection of the search result being a long click-through. For example, a search result having a 7% long click-through rate means that, statistically, for each 100 impressions of the search result, a long click-through will result seven times.
By analysing different users' selections of the search results returned by the information server in response to different search queries, it is possible to determine a “natural” long click-through rate for each position in the list of search results. In this case, the search result ranker compares each identified search result's predicted long click-through rate with the natural long click-through rates at different positions in the list and identifies a position for the search result such that the search result's predicted long click-through rate is higher than the natural long click-through rate at the position immediately following the identified position.
Finally, the search result ranker provides the reordered list of search results to the user through the front-end server (5).