Generating search result summaries

• US 8,032,519 B2
• Filed: 11/16/2010
• Issued: 10/04/2011
• Est. Priority Date: 01/31/2008
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Groeneveld, Dirk, Meyerzon, Dmitriy, Mowatt, David
• Primary Examiner(s): LEROUX, ETIENNE PIERRE
• Application Number:
  US12/947,649
• Publication Number:
  US 20110066611A1
• Time in Patent Office: 322 Days
• Field of Search: 707/723, 707/727, 707/758, 707/713
• US Class Current: 707/713
• CPC Class Codes: G06F 16/345 Summarisation for human users
• CPC Class Codes: G06F 16/9535 Search customisation based on user profiles and personalisation

Abstract

Embodiments are configured to provide a summary of information associated with one or more search results. In an embodiment, a system includes a summary generator that can be configured to provide a summary of information including one or more snippets associated with a search term or search terms. The system includes a ranking component that can be used to rank snippets and the ranked snippets can be used when generating a summary that includes one or more ranked snippets. In one embodiment, the system can be configured to include one or more filters that can be used to filter snippets and the filtered snippets can be used when generating a summary. Other embodiments are available.

Claims

1. A method comprising:

generating a number of snippets based in part on at least one query term;

ranking the number of snippets using a ranking formula to rank each snippet based in part on a number of tunable coefficients, number of words in each snippet, one or more of a number of punctuation symbols and numbers in each snippet, and a score of a proximity feature associated with each snippet, the proximity feature being associated with a snippet span; and

,

providing a result summary based in part on use of the ranking formula.

2. The method of claim 1, further comprising using the ranking formula to rank each snippet using one or more ranking features including a first feature associated with the number of words in each snippet, a second feature associated with the number of punctuation symbols in each snippet, a third feature associated with the number of numbers in each snippet, and a fourth feature associated with the score of the proximity feature.

3. The method of claim 1, further comprising using the ranking formula to rank snippets based in part on the number of words in each snippet multiplied by a first tunable coefficient, the number of punctuation symbols in each snippet multiplied by a second tunable coefficient, the number of numbers in each snippet multiplied by a third tunable coefficient, and the score of the proximity feature multiplied by a fourth tunable coefficient.

4. The method of claim 3, further comprising determining the score of the proximity feature according to the snippet span generated using a number of query terms in each span and number of terms in each span.

5. The method of claim 1, further comprising calculating the score of a span for the proximity feature based in part on a tunable proximity factor, a number of query terms in the span, and a number of terms in the span.

6. The method of claim 1, further comprising using one of a pre-ranking filter and a post-ranking filter to filter snippets as part of reducing redundancy in a search result summary.

7. The method of claim 1, further comprising using at least one pre-ranking filter selected from a query filter and a punctuation filter.

8. The method of claim 7, further comprising using at least one post-ranking filter selected from a title filter and a similarity filter.

9. The method of claim 1, further comprising using a filter to filter snippets including using one or more of a query filter, a punctuation filter, a title filter, and a similarity filter.

10. The method of claim 9, further comprising using the query filter to filter snippets based in part on included query terms.

11. The method of claim 9, further comprising using the punctuation filter to filter snippets based in part on a number and type of punctuation symbols.

12. The method of claim 9, further comprising using the similarity filter to generate an index associated with the snippets including determining whether to discard certain snippets based in part on an examination of the index.

13. A method comprising:

providing snippets based in part on one or more inputs;

filtering the snippets using at least one filter;

ranking the snippets including using a ranking algorithm to rank the snippets based in part on the number of words in each snippet multiplied by a first tunable coefficient, the number of punctuation symbols in each snippet multiplied by a second tunable coefficient, and a proximity feature multiplied by a third tunable coefficient; and

,

generating a search summary using one or more ranked snippets.

14. The method of claim 13, further comprising using a starvation handling procedure to complete a summary character limitation.

15. The method of claim 13, further comprising filtering the snippets using one or more pre-ranking filters and one or more post-ranking filters.

16. The method of claim 15, further comprising filtering the snippets using the one or more pre-ranking filters and one or more post-ranking filters selected from a query filter, a punctuation filter, a title filter, and a similarity filter.

17. The method of claim 13, further comprising filtering ranked snippets using an indexed list in part to determine use of one or more ranked snippets as part of providing a search result summary.

18. A system comprising:

a snippet generator to generate snippets based in part on one or more inputs and one or more chunks of information;

a ranking component to rank snippets and generate ranked snippets using a ranking algorithm to rank each snippet based in part on a determination using a number of words in each snippet and a first tunable coefficient, a number of numbers in each snippet and a second tunable coefficient, a number of punctuation symbols and a third tunable coefficient, and a score of a proximity feature and a fourth tunable coefficient, wherein one or more ranked snippets are used to provide a search result summary; and

,

storage to store a search result summary.

19. The system of claim 18, wherein the ranking component ranks the snippets using a ranking algorithm to calculate a rank equal to:

w_sc_w+p_sc_p+P_sc_p+n_sc_n+r_sc_r,

wherein,

w_s represents the number of words in each snippet,

p_s represents a number of punctuation symbols in each snippet,

P_s represents a subset of p_s, which corresponds to particular punctuation symbols in each snippet,

n_s represents the number of numbers in each snippet,

r_s represents the score of the proximity feature associated with each snippet, and

c_w, c_p, c_n, and c_r are tunable coefficients.

20. The system of claim 19, wherein the ranking component uses the proximity feature and a score of a span equal to:

f×

2^2t-l

wherein,

f is a tunable proximity factor,

t is a number of query terms in the span, and

l is a number of terms in the span.