INTEGRATING EXTERNAL RELATED PHRASE INFORMATION INTO A PHRASE-BASED INDEXING INFORMATION RETRIEVAL SYSTEM

• US 20090070312A1
• Filed: 09/07/2007
• Published: 03/12/2009
• Est. Priority Date: 09/07/2007
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Patterson, Anna L.
• Granted Patent: US 8,117,223 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

An information retrieval system uses phrases to index, retrieve, organize and describe documents, analyzing documents and storing the results of the analysis as phrase data. Phrases are identified that predict the presence of other phrases in documents. Documents are the indexed according to their included phrases. Related phrases and phrase extensions are also identified. Changes to existing phrase data about a document collection submitted by a user is captured and analyzed, and the existing phrase data is updated to reflect the additional knowledge gained through the analysis.

Claims

1. A method for updating phrases associated with a limited document collection, comprising:

determining a list of top phrases for the limited document collection, at least in part based on presence of related phrases of the top phrases;

receiving a replacement phrase for at least one of the top phrases; and

updating related phrase data for the replacement phrase from the related phrase data of the top phrase that is being replaced.

2. A method of determining top phrases of a limited document collection, comprising:

determining top phrases for each of a plurality of documents in the limited document collection, at least in part based on presence of related phrases of the top phrases in each document, each top phrase being associated with a score;

for each top phrase of a document, determining an aggregate score for the top phrase corresponding to the top phrase'"'"'s scores for documents in which it appears in the limited document collection;

selecting a set of top phrases with the highest aggregate scores.

3. The method of claim 2, wherein determining the top phrases of a document comprises:

identifying each phrase of the document; and

for each identified phrase, determining an importance value based on occurrences of related phrases of the identified phrase that are also in the document.

4. The method of claim 3, wherein determining the related phrases of an identified phrase comprises examining a posting list for the identified phrase.

5. The method of claim 3, wherein determining the importance value of an identified phrase is based on a frequency of the related phrases in the document.

6. The method of claim 2, wherein determining an aggregate score for the top phrase comprises summing the scores of the top phrase, for each document for which it is a top phrase.

7. The method of claim 6, further comprising weighting each summed score based on a location within the document collection of the document to which the summed score corresponds.

8. The method of claim 2, further comprising modifying the aggregate score by combining it with a previously determined aggregate score.

9. The method of claim 8, wherein the combination of the aggregate score and the previously determined aggregate score is accomplished by a linear combination of the two scores.

10. A method of updating existing phrase information, responsive to a user requesting a change of a current top phrase for a limited document collection to a replacement top phrase, the method comprising:

associating the replacement top phrase with a root document of the document collection;

associating the current top phrase and replacement top phrase with each other;

adding to phrase information for the replacement top phrase, phrase information for the current top phrase; and

adding to related phrase information of the replacement top phrase, related phrase information of the current top phrase.

11. The method of claim 10, wherein associating the replacement top phrase with the root document of the document collection comprises adding the root document of the document collection to a posting list for the replacement top phrase.

12. The method of claim 10, wherein associating the current top phrase and replacement top phrase with each other comprises:

adding the current top phrase to the related phrase list of the replacement top phrase; and

adding the replacement top phrase to the related phrase list of the current top phrase.

13. The method of claim 10, wherein adding to phrase information for the replacement top phrase, phrase information for the current top phrase, comprises:

determining a count corresponding to a related phrase of the current top phrase that is also a top phrase of the limited document collection; and

creating for the replacement top phrase a corresponding copy of the count;

determining each related phrase of the replacement top phrase that is also a top phrase of the limited document collection; and

updating for the replacement top phrase, a count of the determined related phrases based on a number of occurrences of the related phrases.

14. The method of claim 10, wherein adding to related phrase information for the replacement top phrase, related phrase information of the current top phrase, comprises incrementing counts of related phrases of the replacement top phrase that are also related phrases of the current top phrase.

15. The method of claim 10, further comprising denying the request to change the current top phrase to the existing top phrase, responsive to the current and replacement top phrases having no related phrases in common.

16. The method of claim 10, further comprising decrementing the counts of the related phrases of the replacement top phrase, responsive to the current and replacement top phrases having no related phrases in common.