초록 ▼

A computer system presents retrieved documents to a user, with documents most similar to documents in which the user previously showed high interest being ranked higher than other retrieved documents. The system compares a query vector with feedback query vectors, where each feedback query vector i

A computer system presents retrieved documents to a user, with documents most similar to documents in which the user previously showed high interest being ranked higher than other retrieved documents. The system compares a query vector with feedback query vectors, where each feedback query vector is associated with at least one user feedback vector, and each user feedback vector indicates an aggregate user interest in documents including terms associated with the user feedback vector. The system determines a feedback query vector that is most similar to the query vector, compares the document vectors with a user feedback vector associated with the most similar feedback query vector, and determines the document vector that is most similar to such user feedback vector. The document associated with the most similar document vector is ranked higher than the remaining retrieved documents, when presented to the user.

대표청구항 ▼

What is claimed is: 1. A method of presenting to a user documents retrieved from a document collection by a plurality of search engines in response to a query, each document represented by a document vector in a vector space derived from terms in the document collection, the method comprising: gene

What is claimed is: 1. A method of presenting to a user documents retrieved from a document collection by a plurality of search engines in response to a query, each document represented by a document vector in a vector space derived from terms in the document collection, the method comprising: generating from the query a query vector in the vector space; comparing the query vector with a plurality of feedback query vectors, each feedback query vector associated with at least one user feedback vector, each user feedback vector indicating an aggregated user interest in documents including terms associated with the user feedback vector; determining a feedback query vector that is most similar to the query vector; determining at least one document vector that is most similar to the user feedback vector associated with the most similar feedback query vector; ranking the retrieved documents by assigning the document associated with the at least one most similar document vector to a priority higher than the priority of remaining retrieved documents; and presenting the retrieved documents to the user according to the ranking. 2. The method of claim 1, wherein: the user feedback vector comprises a first user feedback vector and a second user feedback vector; and determining at least one most similar document vector comprises determining at least one first similar document vector that is most similar to the first user feedback vector and at least one second similar document vector that is most similar to the second user feedback vector. 3. The method of claim 2, wherein: the first user feedback vector indicates the user's highest aggregate interest in documents including terms associated with the first user feedback vector; the second user feedback vector indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback vector; and ranking the retrieved documents comprises ranking the document associated with the at least one first similar document vector at a first highest priority and ranking the document associated with the at least one second similar document vector at a second highest priority. 4. The method of claim 3, wherein the first user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the full text and the second user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the summary. 5. The method of claim 1, wherein the retrieved documents have local rankings provided by the search engines, and the method further comprises: normalizing the local rankings associated with the documents; and ranking the other retrieved documents according to their associated normalized local rankings. 6. The method of claim 1, wherein determining a most similar feedback query vector comprises: determining the feedback query vector closest to the query vector and within a predetermined threshold. 7. The method of claim 1, wherein determining at least one most similar document vector comprises: determining at least one document vector closest to the user feedback vector and within a predetermined threshold. 8. A system for presenting to a user documents retrieved from a document collection by a plurality of search engines in response to a query, the documents represented by document vectors in a vector space derived from terms in the document collection, the system comprising: a user feedback database storing feedback query vectors and user feedback vectors associated with the feedback query vectors, the user feedback vectors indicating the user's aggregate interests in selected documents including terms associated with the user feedback vectors, the selected documents corresponding to one or more of the documents retrieved in response to previous queries corresponding to the feedback query vectors; and a processor module coupled to the user feedback database, wherein the processor module: generates a query vector based upon the query; compares the query vector with the feedback query vectors stored in the user feedback; determines a feedback query vector that is most similar to the query vector; compares the document vectors with at least one user feedback vector associated with the most similar feedback query vector; and determines a document vector that is most similar to the user feedback vector; ranks the retrieved documents by assigning the document associated with the most similar document vector to a priority higher than the priority of remaining retrieved documents; and presents the retrieved documents to the user according to the ranking. 9. The system of claim 8, wherein: the user feedback vector comprises a first user feedback vector and a second user feedback vector; the processor module compares the document vector with the first user feedback vector and the second user feedback vector; and the processor module determines a first similar document vector that is most similar to the first user feedback vector and a second similar document vector that is most similar to the second user feedback vector. 10. The system of claim 9, wherein the first user feedback vector indicates the user's highest aggregate interest in documents including terms associated with the first user feedback vector and the second user feedback vector indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback vector. 11. The system of claim 9, wherein the first user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the full text and the second user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the summary. 12. The system of claim 8, wherein the retrieved documents have local rankings provided by the search engines, and the processor module normalizes the local rankings associated with the documents and further ranks the other retrieved documents according to their associated normalized local rankings. 13. The system of claim 8, wherein the processor module compares the query vector with the feedback query vectors by determining the distance between the query vector and the feedback query vector. 14. The system of claim 13, wherein the processor module determines the most similar feedback query vector by determining the feedback query vector closest to the query vector and within a predetermined threshold. 15. The system of claim 8, wherein the processor module compares the document vectors with at least one user feedback vector by determining the distance between the document vectors and the user feedback vector. 16. The system of claim 15, wherein the processor module determines a most similar document vector by determining the document vector closest to the user feedback vector and within a predetermined threshold. 17. A computer program product stored on a computer readable medium and adapted to perform a method of presenting to a user documents retrieved from a document collection by a plurality of search engines in response to a query, each document represented by a document vector in a vector space derived from terms in the document collection, the method comprising: generating from the query a query vector in the vector space; comparing the query vector with a plurality of feedback query vectors, each feedback query vector associated with at least one user feedback vector, each user feedback vector indicating an aggregated user interest in documents including terms associated with the user feedback vector; determining a feedback query vector that is most similar to the query vector; determining at least one document vector that is most similar to the user feedback vector associated with the most similar feedback query vector; ranking the retrieved documents by assigning the document associated with the at least one most similar document vector to a priority higher than the priority of remaining retrieved documents; and presenting the retrieved documents to the user according to the ranking. 18. The computer program product of claim 17, wherein: the user feedback vector comprises a first user feedback vector and a second user feedback vector; and determining at least one most similar document vector comprises determining a first similar document vector that is most similar to the first user feedback vector and a second similar document vector that is most similar to the second user feedback vector. 19. The computer program product of claim 18, wherein: the first user feedback vector indicates the user's highest aggregate interest in documents including terms associated with the first user feedback vector; the second user feedback vector indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback vector; and ranking the retrieved documents comprises ranking the document associated with the first similar document vector at a first highest priority and ranking the document associated with the second similar document vector at a second highest priority. 20. The computer program product of claim 18, wherein the first user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the full text and the second user feedback vector represents the average of document vectors corresponding to documents of which the user reviewed the summary. 21. The computer program product of claim 17, wherein the retrieved documents have local rankings provided by the search engines, and the method further comprises: normalizing the local rankings associated with the documents; and ranking the other retrieved documents according to their associated normalized local rankings. 22. The computer program product of claim 17, wherein determining a most similar feedback query vector comprises: determining the feedback query vector closest to the query vector and within a predetermined threshold. 23. The computer program product of claim 17, wherein determining at least one most similar document vector comprises: determining at least one document vector closest to the user feedback vector and within a predetermined threshold. 24. A method of selecting documents retrieved from a document collection by a plurality of a search engines in response to a query, the method comprising: receiving a query; comparing the received query with a plurality of feedback queries, each feedback query associated with at least one user feedback data, each user feedback data indicating an aggregated user interest in documents including terms associated with the user feedback; determining a feedback query that is most similar to the received query; determining at least one document that is most similar to the user feedback data associated with the most similar feedback query; selecting the at least one most similar document; ranking the retrieved documents by assigning the at least one most similar document to a priority higher than the priority of remaining retrieved documents; and presenting the retrieved documents to the user according to the ranking. 25. The method of claim 24, wherein: the user feedback data comprises a first user feedback data and a second user feedback data; and determining at least one most similar document comprises determining at least one first similar document that is most similar to the first user feedback data and at least one second similar document that is most similar to the second user feedback data. 26. The method of claim 25, wherein: the first user feedback data indicates the user's highest aggregate interest in documents including terms associated with the first user feedback data; the second user feedback data indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback data; and selecting the at least one most similar document comprises selecting the at least one first document with a first highest priority and selecting the at least one second document with a second highest priority. 27. The method of claim 26, wherein the first user feedback data corresponds to documents of which the user reviewed the full text and the second user feedback data corresponds to documents of which the user reviewed the summary. 28. A system for selecting documents retrieved from a document collection by a plurality of search engines in response to a query the system comprising: a user feedback database storing feedback queries and user feedback data associated with the feedback queries, the user feedback data indicating the user's aggregate interests in selected documents including terms associated with the user feedback data, the selected documents corresponding to one or more of the documents retrieved in response to previous queries corresponding to the feedback queries; and a processor module coupled to the user feedback database, wherein the processor module: receives a query; compares the received query with a plurality of feedback queries, each feedback query associated with at least one user feedback data, each user feedback data indicating an aggregated user interest in documents including terms associated with the user feedback; determines a feedback query that is most similar to the received query; determines at least one document that is most similar to the user feedback data associated with the most similar feedback query; selects the at least one most similar document; ranks the retrieved documents by assigning the at least one most similar document to a Priority higher than the Priority of remaining retrieved documents; and presenting the retrieved documents to the user according to the ranking. 29. The system of claim 28, wherein: the user feedback data comprises a first user feedback data and a second user feedback data; and determining at least one most similar document comprises determining at least one first similar document that is most similar to the first user feedback data and at least one second similar document that is most similar to the second user feedback data. 30. The system of claim 29, wherein: the first user feedback data indicates the user's highest aggregate interest in documents including terms associated with the first user feedback data; the second user feedback data indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback data; and selecting the at least one most similar document comprises selecting the at least one first document with a first highest priority and selecting the at least one second document with a second highest priority. 31. The system of claim 30, wherein the first user feedback data corresponds to documents of which the user reviewed the full text and the second user feedback data corresponds to documents of which the user reviewed the summary. 32. A computer program product stored on a computer readable storage medium and adapted to perform a method of selecting documents retrieved from a document collection by a plurality of search engines in response to a query, the method comprising: receiving a query; comparing the received query with a plurality of feedback queries, each feedback query associated with at least one user feedback data, each user feedback data indicating an aggregated user interest in documents including terms associated with the user feedback; determining a feedback query that is most similar to the received query; determining at least one document that is most similar to the user feedback data associated with the most similar feedback query; selecting the at least one most similar document; ranking the retrieved documents by assigning the at least one most similar document to a Priority higher than the Priority of remaining retrieved documents; and presenting the retrieved documents to the user according to the ranking. 33. The computer program product of claim 32, wherein: the user feedback data comprises a first user feedback data and a second user feedback data; and determining at least one most similar document comprises determining at least one first similar document that is most similar to the first user feedback data and at least one second similar document that is most similar to the second user feedback data. 34. The computer program product of claim 33, wherein: the first user feedback data indicates the user's highest aggregate interest in documents including terms associated with the first user feedback data; the second user feedback data indicates the user's second highest aggregate interest in documents including terms associated with the second user feedback data; and selecting the at least one most similar document comprises selecting the at least one first document with a first highest priority and selecting the at least one second document with a second highest priority. 35. The computer program product of claim 34, wherein the first user feedback data corresponds to documents of which the user reviewed the full text and the second user feedback data corresponds to documents of which the user reviewed the summary.