Document ranking with sub-query series
Systems, methods, media, and other embodiments associated with ranking documents by providing a search engine with a series of sub-queries generated from an original query are described. One example system includes input logic for receiving a query. The example system may include a relaxation logic configured to produce sub-queries from the query. The sub-queries may describe metadata string matching, content string matching, and/or metadata numerical attribute analysis. The sub-queries may be provided by an output logic to a search engine in an order that facilitates defining document relevance without requiring post-retrieval relevance ranking.
1. A computer implemented method, comprising:
receiving a query, the query being configured to cause a search logic to retrieve one or more documents, the documents comprising text elements and metadata elements;
creating from the query a series of sub-queries {SQ 1 , . . . SQ N }, N being an integer greater than 1, the series of sub-queries including a sub-query based on one or more of a numeric metadata link score and a numeric metadata URL depth;
the series of sub-queries being configured so that a sub-query SQ X will cause the search logic to retrieve fewer documents than a sub-query SQ Y when X<Y;
the series of sub-queries being configured so that a sub-query SQ A will cause the search logic to retrieve documents having a higher relevancy than a sub-query SQ B when A<B; and
providing, in order according to the configuration of the series of sub-queries, one or more sub-queries from the series of sub-queries to the search logic until a pre-determined number of documents are retrieved by the search logic.
2. The method of claim 1 , where the series of sub-queries includes a metadata-centric sub-query and further where the metadata-centric sub-query comprises one or more of, a sub-query that matches terms in the query to one or more metadata elements on a phrase level, a sub-query that matches terms in the query to one or more metadata elements on a proximity level, and a sub-query that matches terms in the query to one or more metadata elements on a Boolean AND level.
3. The method of claim 2 , where the metadata elements include one or more of, a title, a description, a keyword, a subject, an author, a uniform resource locator (URL), an H1 headline, an H2 headline, and an anchor text.
4. The method of claim 2 , where the metadata-centric sub-query includes at least one metadata element associated with a metadata element weight, and where the document relevance is a function of:
a type of match between terms in the query and the metadata element.
5. The method of claim 1 , where the series of sub-queries includes a linguistic-centric sub-query and further where the linguistic-centric sub-query includes one or more of, a sub-query that matches terms in the query to one or more text elements on a phrase level, a sub-query that matches terms in the query to one or more text elements on a proximity level, and a sub-query that matches terms in the query to one or more text elements on a Boolean AND level.
6. The method of claim 1 , including creating a sub-query by stemming terms in the query.
7. The method of claim 1 , where the series of sub-queries includes a metadata-centric sub-query and a linguistic-centric sub-query, ordered:
to make matches between terms in the query and metadata elements more determinative of document relevancy than matches between terms in the query and text elements;
to make a phrase match more determinative of document relevancy than a proximity match; and
to make a proximity match more determinative of document relevancy than a Boolean AND match.
8. The method of claim 1 , where document relevance is a function of a metadata element weight and further including establishing the following metadata element to metadata element weight pairs: (title, w 1 ), (description, w 2 ), (keyword, w 3 ), (subject, w 4 ), (author, w 5 ), (url, w 6 ), (H1 headline, w 7 ), (H2 headline, w 8 ), and (anchortext, w 9 ), where w 1 through w 9 are configurable to facilitate determining the document relevance before document retrieval.
9. The method of claim 8 , where w 1 >w 2 >w 3 >w 4 >w 5 >w 6 >w 7 >w 8 >w 9 .
10. The method of claim 1 , where the numeric metadata link score is directly proportional to document relevance and where the numeric metadata URL depth is inversely proportional to document relevance.
11. The method of claim 1 , further comprising
providing retrieved documents to a user, the documents being ordered based on relevance without performing post-retrieval relevance ranking and where duplicate documents are removed.
12. A computer implemented method, comprising:
receiving a query, the query being configured to cause a search logic to retrieve one or more documents, the documents comprising text elements and metadata elements;
creating from the query a series of sub-queries {SQ 1 , . . . SQ N }, N being an integer greater than 1, the series of sub-queries including a sub-query based on one or more of, a numeric metadata link score, and a numeric metadata URL depth, where the numeric metadata link score is directly proportional to document relevance and where the numeric metadata URL depth is inversely proportional to document relevance;
the series of sub-queries being configured so that a sub-query SQ X will cause the search logic to retrieve fewer documents than a sub-query SQ Y when X<Y;
the series of sub-queries being configured so that a sub-query SQ A will cause the search logic to retrieve documents having a higher relevancy than a sub-query SQ B when A<B; and
providing, in order according to the configuration of the series of sub-queries, one or more sub-queries from the series of sub-queries to the search logic until a pre-determined number of documents are retrieved by the search logic.
13. A non-transitory computer-readable storage medium storing processor executable instructions operable to perform a method, the method comprising:
receiving a query, the query being configured to cause a search logic to retrieve one or more documents, the documents comprising text elements and metadata elements;
creating from the query a series of sub-queries {SQ 1 , . . . SQ N }, N being an integer greater than 1, the series of sub-queries including one or more of, a sub-query based on one or more of, a numeric metadata link score, and a numeric metadata URL depth, where the numeric metadata link score is directly proportional to document relevance and where the numeric metadata URL depth is inversely proportional to document relevance;
the series of sub-queries being configured so that a sub-query SQ X will cause the search logic to retrieve fewer documents than a sub-query SQ Y when X<Y;
the series of sub-queries being configured so that a sub-query SQA will cause the search logic to retrieve documents having a higher relevancy than a sub-query SQB when A<B; and
providing, in order providing, in order one or more sub-queries from the series of sub-queries, one or more sub-queries from the series of sub-queries to the search logic until a pre-determined number of documents are retrieved by the search logic.
14. The non-transitory computer-readable medium of claim 13 , where the series of sub-queries includes a metadata-centric sub-query and further where the metadata-centric sub-query comprises one or more of, a sub-query that matches terms in the query to one or more metadata elements on a phrase level, a sub-query that matches terms in the query to one or more metadata elements on a proximity level, and a sub-query that matches terms in the query to one or more metadata elements on a Boolean AND level.
15. The non-transitory computer-readable medium of claim 14 , where the metadata elements include one or more of, a title, a description, a keyword, a subject, an author, a uniform resource locator (URL), an H1 headline, an H2 headline, and an anchor text.
16. The non-transitory computer-readable medium of claim 14 , where the metadata-centric sub-query includes at least one metadata element associated with a metadata element weight, and where the document relevance is a function of a type of match between terms in the query and the metadata element.
17. The non-transitory computer-readable medium of claim 13 , where the series of sub-queries includes a linguistic-centric sub-query and further where the linguistic-centric sub-query includes one or more of, a sub-query that matches terms in the query to one or more text elements on a phrase level, a sub-query that matches terms in the query to one or more text elements on a proximity level, and a sub-query that matches terms in the query to one or more text elements on a Boolean AND level.
18. The non-transitory computer-readable medium of claim 13 , where the instructions include creating a sub-query by stemming terms in the query.
19. The non-transitory computer-readable medium of claim 13 , where the series of sub-queries includes a metadata-centric sub-query and a linguistic-centric sub-query, ordered:
to make matches between terms in the query and metadata elements more determinative of document relevancy than matches between terms in the query and text elements;
to make a phrase match more determinative of document relevancy than a proximity match; and
to make a proximity match more determinative of document relevancy than a Boolean AND match.
20. The non-transitory computer-readable medium of claim 13 , where document relevance is a function of a metadata element weight and further including establishing the following metadata element to metadata element weight pairs: (title, w 1 ), (description, w 2 ), (keyword, w 3 ), (subject, w 4 ), (author, w 5 ), (url, w 6 ), (H1 headline, w 7 ), (H2 headline, w 8 ), and (anchortext, w 9 ), where w 1 through w 9 are configurable to facilitate determining the document relevance before document retrieval.
21. The non-transitory computer-readable medium of claim 20 , where w 1 >w 2 >w 3 >w 4 >w 5 >w 6 >w 7 >w 8 >w 9 .
22. The non-transitory computer-readable medium of claim 13 , comprising providing retrieved documents to a user, the documents being ordered based on relevance without performing post-retrieval relevance ranking and where duplicate documents are removed.