Method and system for quantifying the quality of search results based on cohesion
A method and system for quantifying the quality of search results from a search engine based on cohesion. The method and system include modeling a set of search engine search results as a cluster and measuring the cohesion of the cluster. In an embodiment, the cohesion of the cluster is the average similarity between the cluster elements to a centroid vector. The centroid vector is the average of the weights of the vectors of the cluster. The similarity between the centroid vector and the cluster's elements is the cosine similarity measure. Each document in the set of search results is represented by a vector where each cell of the vector represents a stemmed word. Each cell has a cell value which is the frequency of the corresponding stemmed word in a document multiplied by a weight that takes into account the location of the stemmed word within the document.
1. A computer-implemented method for evaluating a set of search results associated with a search engine, the method comprising:
representing the set of search results associated with the search engine as a set of vectors, each vector in the set of vectors comprising one or more cells, each cell of the one or more cells of the vector having a cell value that corresponds to the frequency of a word from the set of search results in a search result that corresponds to the vector;
computing a centroid vector based on the set of vectors, wherein said centroid vector represents a prototype search result;
computing a cohesion metric, based on the set of vectors and the centroid vector, that quantifies the quality of the set of search results, wherein said cohesion metric represents the similarity between the set of search results to the prototype search result; and
storing the cohesion metric in volatile or non-volatile memory.
2. The computer-implemented method of claim 1 , wherein the set of search results is associated with a set of documents, and wherein each vector in the set of vectors represents a document in the set of documents.
3. The computer-implemented method of claim 2 , wherein each of the one or more cells contained in each vector represents a word in the set of documents, each of the one or more cells having a cell value that represents, at least in part, the frequency of the word in a document in the set of documents.
4. The computer-implemented method of claim 3 , wherein the word is a stemmed word.
5. The computer-implemented method of claim 1 , wherein computing the centroid vector includes computing the average of the weights of the vectors in the set of vectors.
6. The computer-implemented method of claim 1 , wherein computing the cohesion metric includes computing the average of the similarities between each vector in the set of vectors and the centroid vector.
7. The computer-implemented method of claim 6 , wherein the similarity between the centroid vector and a vector in the set of vectors includes computing the cosine similarity of the centroid vector and the vector in the set of vectors.
8. The computer-implemented method of claim 1 , further comprising:
determining one or more search engines capable of responding to a search query, wherein at least one of the search engines is the search engine associated with the set of search results;
selecting, based on the cohesion metric, the search engine associated with the search results to respond to the search query; and
sending the search query to the search engine associated with set of search results.
9. The computer-implemented method of claim 8 , wherein the search query contains one or more search terms, and wherein the determining the one or more search engines capable of responding to the search query includes determining, based on statistics of the one or more search terms stored in a database, the one or more search engines that are capable of responding to the search query.
10. A computer-implemented method for generating a set of meta-search results comprising, for each set of search results in a plurality of sets of search results:
(a) representing the set of search results as a set of vectors, each vector in the set of vectors comprising one or more cells, each cell of the one or more cells of the vector having a cell value that corresponds to the frequency of a word from the set of search results in a search result that corresponds to the vector;
(b) computing a centroid vector based on the set of vectors, wherein said centroid vector represents a prototype search result,
(c) computing a cohesion metric, based on the set of vectors and the centroid vector, that quantifies the quality of the set of search results, wherein said cohesion metric represents the similarity between the set of search results to the prototype search result, and
(d) storing the cohesion metric in volatile or non-volatile memory; generating a set of meta-search results comprising one or more search results from the sets of search results based on the cohesion metric computed for each set of search results.
11. The computer-implemented method of claim 10 , wherein each set of search results in the plurality of sets of search results is associated with a different one of a plurality of search engines.
12. The computer-implemented method of claim 10 , wherein the steps (a), (b), (c), and (d) performed for each set of search results in the plurality of sets of search results are performed in response to receiving the plurality of sets of search results from a plurality of search engines.
13. A machine-readable storage medium storing one or more sequences of instructions which, when executed by one or more processors, causes the one or more processors to perform the steps of:
representing a set of search results associated with a search engine as a set of vectors, each vector in the set of vectors comprising one or more cells, each cell of the one or more cells of the vector having a cell value that corresponds to the frequency of a word from the set of search results in a search result that corresponds to the vector;
computing a centroid vector based on the set of vectors, wherein said centroid vector represents a prototype search result;
computing a cohesion metric, based on the set of vectors and the centroid vector, that quantifies the quality of the set of search results, wherein said cohesion metric represents the similarity between the set of search results to the prototype search result; and
storing the cohesion metric in volatile or non-volatile memory.
14. The machine-readable storage medium of claim 13 , wherein the set of search results is associated with a set of documents, and wherein each vector in the set of vectors represents a document in the set of documents.
15. The machine-readable storage medium of claim 14 , wherein each of the one or more cells contained in each vector represents a word in the set of documents, each of the one or more cells having a cell value that represents, at least in part, the frequency of the word in a document in the set of documents.
16. The machine-readable storage medium of claim 15 , wherein the word is a stemmed word.
17. The machine-readable storage medium of claim 13 , wherein computing the centroid vector includes computing the average of the weights of the vectors in the set of vectors.
18. The machine-readable storage medium of claim 13 , wherein computing the cohesion metric includes computing the average of the similarities between each vector in the set of vectors and the centroid vector.
19. The machine-readable storage medium of claim 18 , wherein the similarity between the centroid vector that represents a prototype search result and a vector in the set of vectors includes computing the cosine similarity of the centroid vector and the vector in the set of vectors.
20. The machine-readable storage medium of claim 13 , wherein execution of the one or more sequences of instructions by the one or more processors causes the one or more processors to perform the steps of:
determining one or more search engines capable of responding to a search query, wherein at least one of the search engines is the search engine associated with the set of search results;
selecting, based on the cohesion metric that quantifies the quality of the set of search results, the search engine associated with the search results to respond to the search query; and
sending the search query to the search engine associated with set of search results.
21. The machine-readable storage medium of claim 20 , wherein the search query contains one or more search terms, and wherein the determining the one or more search engines capable of responding to the search query includes determining, based on statistics of the one or more search terms stored in a database, the one or more search engines that are capable of responding to the search query.
22. A machine-readable storage medium storing one or more sequences of instructions which, when executed by one or more processors, causes the one or more processors to perform the steps of:
for each set of search results in a plurality of sets of search results:
(a) representing the set of search results as a set of vectors, each vector in the set of vectors comprising one or more cells, each cell of the one or more cells of the vector having a cell value that corresponds to the frequency of a word from the set of search results in a search result that corresponds to the vector;
(b) computing a centroid vector based on the set of vectors, wherein said centroid vector represents a prototype search result,
(c) computing a cohesion metric, based on the set of vectors and the centroid vector, that quantifies the quality of the set of search results, wherein said cohesion metric represents similarity between the set of search results to the prototype search result, and
(d) storing the cohesion metric in volatile or non-volatile memory.
23. The machine-readable storage medium of claim 22 , wherein each set of search results in the plurality of sets of search results is associated with a different one of a plurality of search engines.
24. The machine-readable storage medium of claim 22 , wherein the steps (a), (b), (c), and (d) performed for each set of search results in the plurality of sets of search results are performed in response to receiving the plurality of sets of search results from a plurality of search engines.