Returning search results utilizing topical user click data when search queries are dissimilar
AssigneeInternational Business Machines Corporation
InventorsPriyanka Agrawal, Man Chu, Steven M. Pritko, Justin A. Ziniel
Life of this patent
Estimating the term…
Abstract
For a particular search query that is received by a host system, a preliminary set of search results is generated. The host system maps the query to one or more topics that are representative to the query. The host system provides topical click data associated with the topic to a machine-learning module that determines a relevancy score of the result utilizing the topical click data. The host system re-ranks the order of the result within a set of results based upon the relevancy score.
Claims9
A method of re-ranking search results by utilizing topical click data in an environment where search result queries are ostensibly dissimilar, the method comprising:
- determining, with a host device processor, a preliminary result [R] comprising a plurality of preliminary results of a search query [Q] received by the host device;
- subsequent to determining the preliminary result [R] of the search query [Q], saving a <Q,R> data pair to a storage device of the host device, the <Q,R> data pair consisting of the search query [Q] and the preliminary result [R];
- determining, with the host device processor, a plurality of topics [T1, Tn] that are representative of the search query [Q], wherein the plurality of topics [T1, Tn] are each single word topics, wherein the single word topic [T1] is weighted as being more representative of the search query [Q] relative to the single word topic [Tn];
- subsequent to determining the plurality of topics [T1, Tn] that are representative of the search query [Q], saving a <Q,R,Tn> data triplet to the storage device, the <Q,R,Tn> data triplet consisting of the search query [Q], the preliminary result [R], and the single word topic [Tn];
- retrieving, with the host device processor, topical click data of a single word topic that is most representative of the search query [Q], wherein retrieving topical click data of the single word topic that is most representative of the search query [Q] comprises:
- attempting to retrieve, with the host device processor, topical click data [C1] of the single word topic [T1] that is weighted as more representative of the search query [Q];
- determining, with the host device processor, that no topical click data [C1] of the single word topic [T1] exists;
- and
- subsequent to determining that no topical click data [C1] of the single word topic [T1] exists, retrieving, with the host device processor, topical click data [Cn] of the single word topic [Tn], wherein retrieving topical click data [Cn] of the single word topic [Tn] comprises:
- reading a topical data structure <Q,R,Tn,Cn> from the storage device of the host device, wherein the topical data structure <Q,R,Tn,Cn> consists of the topical click data [Cn], the topic [Tn], the preliminary result [R], and the search query [Q];
- determining, with the host device processor, a relevance value [V] of each of the plurality of preliminary results of the preliminary result [R] utilizing the topical click data [Cn];
- subsequent to determining the relevance value [V] of the result [R], saving a <Q,R,Tn,Cn,V> data quintuple to the storage device, the <Q,R,Tn,Cn,V> data quintuple consisting of the search query [Q], the preliminary result [R], the single word topic [Tn], the topical click data [Cn], and the relevance value [V] of each of the plurality of preliminary results of the preliminary result [R];
- and
- re-ranking, with the host device processor, the order of the plurality of preliminary results of the preliminary result [R] based upon the relevance value [V] of each of the plurality of preliminary results.
A computer program product for re-ranking search results by utilizing topical click data in an environment where search result queries are ostensibly dissimilar, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions readable by the computer to cause the computer to:
- determine a preliminary result [R] comprising a plurality of preliminary results of a search query [Q] received by the host device;
- subsequent to determining the preliminary result [R] of the search query [Q], save a <Q,R> data pair to a storage device of the host device, the <Q,R> data pair consisting of the search query [Q] and the preliminary result [R];
- determine a plurality of topics [T1, Tn] that are representative of the search query [Q], wherein the plurality of topics [T1, Tn] are each single word topics, wherein the single word topic [T1] is weighted as being more representative of the search query [Q] relative to the single word topic [Tn];
- subsequent to determining the plurality of topics [T1, Tn] that are representative of the search query [Q], save a <Q,R,Tn> data triplet to the storage device, the <Q,R,Tn> data triplet consisting of the search query [Q], the preliminary result [R], and the single word topic [Tn];
- retrieve topical click data of a single word topic that is most representative of the search query [Q], wherein the program instructions that cause the computer to retrieve topical click data of the single word topic that is most representative of the search query [Q] further causes the computer to:
- attempt to retrieve topical click data [C1] of the single word topic [T1] that is weighted as more representative of the search query [Q];
- determine that no topical click data [C1] of the single word topic [T1] exists;
- and
- subsequent to the determination that no topical click data [C1] of the single word topic [T1] exists, retrieve topical click data [Cn] of the single word topic [Tn], wherein the retrieval of topical click data [Cn] of the single word topic [Tn] comprises:
- reading a topical data structure <Q,R,Tn,Cn> from the storage device of the host device, wherein the topical data structure <Q,R,Tn,Cn> consists of the topical click data [Cn], the topic [Tn], the preliminary result [R], and the search query [Q];
- determine a relevance value [V] of each of the plurality of preliminary results of the preliminary result [R] utilizing the topical click data [Cn];
- subsequent to determining the relevance value [V] of the result [R], save a <Q,R,Tn,Cn,V> data quintuple to the storage device, the <Q,R,Tn,Cn,V> data quintuple consisting of the search query [Q], the preliminary result [R], the single word topic [Tn], the topical click data [Cn], and the relevance value [V] of each of the plurality of preliminary results of the preliminary result [R];
- and
- re-rank the order of the plurality of preliminary results of the preliminary result [R] based upon the relevance value [V] of each of the plurality of preliminary results.
A computer comprising:
- a processor;
- and
- memory communicatively coupled to the processor, wherein the memory is encoded with instructions, wherein the instructions when executed by the processor cause the processor to:
- determine a preliminary result [R] comprising a plurality of preliminary results of a search query [Q] received by the host device;
- subsequent to determining the preliminary result [R] of the search query [Q], save a <Q,R> data pair to a storage device of the host device, the <Q,R> data pair consisting of the search query [Q] and the preliminary result [R];
- determine a plurality of topics [T1, Tn] that are representative of the search query [Q], wherein the plurality of topics [T1, Tn] are each single word topics, wherein the single word topic [T1] is weighted as being more representative of the search query [Q] relative to the single word topic [Tn];
- subsequent to determining the plurality of topics [T1, Tn] that are representative of the search query [Q], save a <Q,R,Tn> data triplet to the storage device, the <Q,R,Tn> data triplet consisting of the search query [Q], the preliminary result [R], and the single word topic [Tn];
- retrieve topical click data of a single word topic that is most representative of the search query [Q], wherein the program instructions that cause the processor to retrieve topical click data of the single word topic that is most representative of the search query [Q] further causes the processor to:
- attempt to retrieve topical click data [C1] of the single word topic [T1] that is weighted as more representative of the search query [Q];
- determine that no topical click data [C1] of the single word topic [T1] exists;
- and
- subsequent to the determination that no topical click data [C1] of the single word topic [T1] exists, retrieve topical click data [Cn] of the single word topic [Tn], wherein the retrieval of topical click data [Cn] of the single word topic [Tn] comprises:
- reading a topical data structure <Q,R,Tn,Cn> from [a] the storage device of the host device, wherein the topical data structure <Q,R,Tn,Cn> consists of the topical click data [Cn], the topic [Tn], the preliminary result [R], and the search query [Q];
- determine a relevance value [V] of each of the plurality of preliminary results of the preliminary result [R] utilizing the topical click data [Cn];
- subsequent to determining the relevance value [V] of the result [R], save a <Q,R,T,C,V> data quintuple to the storage device, the <Q,R,Tn,Cn,V> data quintuple consisting of the search query [Q], the preliminary result [R], the single word topic [Tn], the topical click data [Cn], and the relevance value [V] of each of the plurality of preliminary results of the preliminary result [R];
- and
- re-rank the order of the plurality of preliminary results of the preliminary result [R] based upon the relevance value [V] of each of the plurality of preliminary results.
Dependencies are inferred from the claim text.