GENERATING SNIPPETS BASED ON CONTENT FEATURES

• US 20110282651A1
• Filed: 05/11/2010
• Published: 11/17/2011
• Est. Priority Date: 05/11/2010
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: TURCHETTO, RICCARDO, PAN, FENG, NYGAARD, VALERIE ROSE, CONVERSE, TIMOTHY MCDONNELL, CHAN, JOANNA MUN YEE, KING, TRACY HOLLOWAY, BURBANK, ANDREA RYERSON, REINHOLD, JAMES MICHAEL, BIEMANN, CHRISTIAN, AHN, DAVID DONGJAH
• Granted Patent: US 8,788,260 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 16/338 Presentation of query results
• CPC Class Codes: G06F 16/345 Summarisation for human users

Abstract

Systems, methods, and computer storage media having computer-executable instructions embodied thereon that facilitate generation of snippets. In embodiments, text features within a keyword-sentence window are identified. The text features are utilized to determine break features that indicate favorability of breaking at a particular location of the keyword-sentence window. The break features are used to recognize features of partial snippets such that a snippet score to indicate the strength of the partial snippet can be calculated. Snippet scores associated with partial snippets are compared to select an optimal snippet, that is, the snippet having the highest snippet score.

Claims

1. One or more computer storage media having computer-executable instructions embodied thereon, that when executed, cause a computing device to perform a method for facilitating generation of snippets provided in association with search results, the method comprising:

referencing a keyword-sentence window comprising a sequence of tokens including one or more keywords that match one or more query terms;

identifying a part-of-speech for one or more tokens within the keyword-sentence window; and

utilizing the part-of-speech corresponding with each of the one or more tokens to identify one or more text features associated with a span including two or more tokens, the one or more text features being used to generate a snippet comprising a portion of the keyword-sentence window truncated at optimal snippet boundaries.

2. The media of claim 1 further comprising displaying the snippet in association with a search result.

3. The media of claim 1, wherein the snippet comprising the portion of the keyword-sentence window truncated at optimal snippet boundaries provides text that breaks at natural beginning and ending points.

4. The media of claim 1, wherein the snippet effectively characterizes a document in association with a search result such that potentially relevant information is not truncated from the keyword-sentence window.

5. The media of claim 1, wherein the one or more text features comprise a bigram type, a named entity, a breakpoint, or a combination thereof.

6. The media of claim 5, wherein the bigram type comprises an indication of a part-of-speech sequence corresponding with a bigram.

7. The media of claim 5, wherein the named entity is identified based on a pattern of part-of-speech identifiers.

8. The media of claim 5, wherein the breakpoint comprises an indication of a point at which the keyword-sentence window can be favorably truncated.

9. The media of claim 1 further comprising identifying one or more additional text features based on tokens or spans within the keyword-sentence window.

10. The media of claim 9 further comprising utilizing the one or more text features and the one or more additional text features to determine one or more break features that indicate favorability of breaking the keyword-sentence window.

11. The media of claim 10, wherein the one or more break features comprise a breaking indicator, a span indicator, or a combination thereof.

12. The method of claim 11, wherein the one or more break features are used to generate the snippet comprising the portion of the keyword-sentence window truncated at the optimal snippet boundaries.

13. A method for facilitating generation of snippets provided in association with search results, the method comprising:

identifying one or more text features for a plurality of spans within a keyword-sentence window, at least a portion of the text features being identified based on a part-of-speech identifier associated with one or more tokens of the span;

determining one or more break features associated with one or more of the plurality of spans using the one or more text features, the one or more break features providing an indication of whether a snippet boundary is favorable relative to a particular position within the keyword-sentence window; and

utilizing the one or more break features to generate a snippet comprising a portion of the keyword-sentence window truncated at appropriate snippet boundaries.

14. The method of claim 13, wherein the one or more break features comprise a breaking indicator to indicate favorability of breaking a sentence and a span indicator to indicate whether a span is interesting or droppable.

15. The method of claim 14 further comprising using the span indicator and the breaking indicator to determine snippet features that indicate features of the snippet.

16. The method of claim 15 further comprising using the snippet features to calculate a snippet score for the snippet that indicates favorability of truncating the keyword-sentence window in accordance with the snippet boundaries.

17. The method of claim 13 further comprising displaying the snippet in association with a search result.

18. One or more computer storage media having computer-executable instructions embodied thereon, that when executed, cause a computing device to perform a method for facilitating generation of snippets provided in association with search results, the method comprising:

identifying one or more text features associated with spans within a keyword-sentence window including at least one keyword that matches at least one query term, at least a portion of the one or more text features being identified based on a part-of-speech identifier associated with one or more tokens of the span;

determining one or more break features associated with the spans using the one or more text features, the one or more break features providing an indication of whether a snippet boundary is favorable relative to a particular position within the keyword-sentence window;

generating a plurality of partial snippets comprising portions of the keyword-sentence window;

for each partial snippet, identifying one or more snippet features that indicate a relative strength of truncating the keyword-sentence window in accordance with the partial snippet;

determining a score for each of the plurality of partial snippets that indicates favorability of truncating the keyword-sentence window at snippet boundaries as indicated in the partial snippet, the score being based on the one or more snippet features; and

based on the scores, selecting a partial snippet from the plurality of partial snippets to display in association with a search result, the selected partial snippet designated as having optimal snippet boundaries.

19. The media of claim 18 further comprising displaying the selected partial snippet in association with a search result.

20. The media of claim 18, wherein generating the plurality of partial snippets comprises expanding one or more previous partial snippets.