Factoid-based searching

• US 20070136280A1
• Filed: 12/13/2005
• Published: 06/14/2007
• Est. Priority Date: 12/13/2005
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Li, Hang, Cao, Yunbo, Gao, Jianfeng
• Granted Patent: US 7,707,204 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06F 16/313 Selection or weighting of terms for indexing

Abstract

A query and a factoid type selection are received from a user. An index of passages, indexed based on factoids, is accessed and passages that are related to the query, and that have the selected factoid type, are retrieved. The retrieved passages are ranked and provided to the user based on a calculated score, in rank order.

Claims

1. A computer implemented method of accessing information, comprising:

receiving a first textual input indicative of information to be accessed;

receiving, in addition to the first textual input, a type input indicative of a type of information to be accessed;

identifying stored information, having the type indicated by the type input, based on the first textual input; and

providing an output indicative of the stored information identified.

2. The method of claim 1 wherein identifying stored information comprises:

accessing an index of the stored information based on the type input and the first textual input.

3. The method of claim 2 wherein receiving the type input comprises:

receiving a factoid type selection input indicative of a factoid type of information to be accessed.

4. The method of claim 3 wherein the index indexes the stored information based on factoid types included in the stored information and wherein accessing the index comprises:

accessing the index based on the factoid type selection input to identify the stored information having factoids of the factoid type indicated by the factoid type selection input.

5. The method of claim 4 wherein the index stores passages generated from the stored information and indexed by factoid type and wherein accessing the index comprises:

retrieving passages from the index having the factoid type indicated by the factoid type selection input.

6. The method of claim 5 wherein identifying stored information comprises:

scoring the retrieved passages based on a relationship between the first textual input and the retrieved passages, and based on pre-computed scores associated with each passage indicative of a relationship between a given passage and a selected factoid type.

7. The method of claim 6 wherein identifying stored information comprises:

ranking the retrieved passages based on the scoring for the retrieved passages.

8. A method of indexing stored information comprising:

identifying expressions in the stored information containing a factoid of a selected factoid type;

constructing a passage corresponding to each expression identified;

computing a score corresponding to each passage based on a relationship between the passage and the selected factoid type; and

generating an index of the passages and corresponding scores based on the factoid type contained in the expression.

9. The method of claim 8 and further comprising:

repeating the steps of identifying, constructing, computing and generating for each of a plurality of factoid types.

10. The method of claim 8 wherein constructing a passage comprises:

for each expression, extracting text from the stored information containing the expression.

11. The method of claim 10 wherein extracting text comprises:

determining whether a paragraph containing a given expression has at least a threshold number of words; and

if so, extracting a sentence containing the given expression and adding the sentence to the passage.

12. The method of claim 11 wherein extracting text further comprises:

if the paragraph containing the given expression does not have at least the threshold number of words, extracting the paragraph containing the expression and adding it to the passage.

13. The method of claim 12 wherein extracting text further comprises:

if the paragraph containing the given expression contains only the given expression, then extracting a paragraph preceding the given expression and a paragraph subsequent to the given expression and adding the previous and subsequent paragraphs to the passage.

14. The method of claim 13 wherein extracting text further comprises:

extracting title information indicative of a title of a document containing the given expression and adding the title information to the passage.

15. The method of claim 13 wherein extracting text further comprises:

if the given expression is contained in a table, extracting caption information indicative of a caption of the table and adding the caption information to the passage.

16. The method of claim 13 wherein extracting text further comprises:

repeatedly extracting paragraphs or sentences and adding them to the passage until the passage contains at least a threshold number of words.

17. The method of claim 8 wherein computing a score corresponding to each passage comprises:

determining a likelihood that a factoid in an expression corresponding to a given passage answers a request represented by at least one content word.

18. The method of claim 17 wherein determining a likelihood comprises:

computing a ranking function for each content word in the given passage.

19. The method of claim 18 wherein computing a ranking function comprises:

linearly combining weighted features in a multi-dimensional feature vector for each content word.

20. A data accessing system, comprising:

an indexing system configured to index passages generated from a data corpus, comprising data from one or more sources, based on a plurality of pre-defined factoid types, the indexing system being further configured to index a score corresponding to each passage, the score being indicative of a strength of relationship between the passage and a factoid type; and

a runtime system configured to receive a query and a factoid type selection and retrieve passages indexed by the indexing system that are related to the query and of a type indicated by the factoid type selection.