Adaptive and personalized tag recommendation
AssigneeInternational Business Machines Corporation
InventorsHongxia Jin, Qihua Wang
Life of this patent
Estimating the term…
Abstract
Embodiments of the invention relate to techniques for personalized tag recommendation for enterprise social bookmarking systems. According to one embodiment of the invention, a method comprises computing at least one set of candidate tags for a document in accordance with at least one factor, and combining the at least one set of candidate tags from different factors into an overall recommendation list of candidate tags. At least one suggested tag from the overall recommendation list is returned as a result of the combination. At least one applied tag that is applied to the document by a user of a computer system is recorded, and the combination is adjusted based on the at least one applied tag.
Claims23
A method, comprising:
- computing, by a processor, a plurality of sets of candidate tags for a document, wherein the plurality of sets of candidate tags include a set of candidate tags based on document-specific factors including keywords in the document, a set of candidate tags based on collaborative factors including tags that other users have applied to the document, a set of candidate tags based on personalized factors including information about a user and a set of candidate tags based on tag co-occurrences factors including tags that are often applied together with certain keywords;
- combining the plurality of sets of candidate tags from different factors into an overall recommendation list of candidate tags;
- returning at least one suggested tag from the overall recommendation list of candidate tags;
- recording at least one applied tag that is applied to the document by a user of a computer system;
- and
- adjusting the overall recommendation list of candidate tags based on the at least one applied tag.
A method, comprising:
- computing, by a processor, a plurality of sets of candidate tags for a document in accordance with different factors by extracting at least one keyword from the document, and associating an importance score with each keyword, wherein each keyword is a candidate tag, wherein the plurality of sets of candidate tags include a set of candidate tags based on document-specific factors including keywords in the document, a set of candidate tags based on collaborative factors including tags that other users have applied to the document, a set of candidate tags based on personalized factors including information about a user and a set of candidate tags based on tag co-occurrences factors including tads that are often applied together with certain keywords;
- combining the plurality of sets of candidate tags from different factors into an overall recommendation list of candidate tags;
- returning at least one suggested tag from the overall recommendation list of candidate tags;
- recording at least one applied tag that is applied to the document by a user of a computer system;
- and
- adjusting the overall recommendation list of candidate tagsbased on the at least one applied tag.
A computer program product comprising a non-transitory computer readable storage medium having computer readable program code embodied therewith, said computer readable program code comprising computer readable program code configured to:
- compute a plurality of sets of candidate tags for a wherein each of the plurality of sets of candidate tags is based on different factors, wherein the plurality of sets of candidate tags include a set of candidate tags based on document-s ecific factors including keywords in the document, a set of candidate tags based on collaborative factors including tags that other users have applied to the document, a set of candidate tags based on personalized factors including information about a user and a set of candidate to s based on tag co-occurrences factors including tags that are often applied together with certain keywords;
- combine the plurality of sets of candidate tags from different factors into an overall recommendation list of candidate tags;
- return at least one suggested tag from the overall recommendation list of candidate tags;
- record at least one applied tag that is applied to the document by a user of a computer system;
- and
- adjust the overall recommendation list of candidate tags based on the at least one applied tag.
Dependencies are inferred from the claim text.