Voice-based search processing

• US 8,260,809 B2
• Filed: 06/28/2007
• Issued: 09/04/2012
• Est. Priority Date: 06/28/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Platt, John C., Flake, Gary W., Naam, Ramez, Gupta, Anoop, Hurst-Hiller, Oliver, Griffin, Trenholme J.
• Primary Examiner(s): Saeed, Usmaan
• Assistant Examiner(s): Ahluwalia, Navneet K
• Application Number:
  US11/770,535
• Publication Number:
  US 20090006345A1
• Time in Patent Office: 1,895 Days
• Field of Search: 707 1- 10, 707100-1041, 707/110, 707/771, 707/784
• US Class Current: 707/771
• CPC Class Codes: G06F 16/90332 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 40/274 Converting codes to words; Guess-ahead of partial word inputs
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Architecture for completing search queries by using artificial intelligence based schemes to infer search intentions of users. Partial queries are completed dynamically in real time. Additionally, search aliasing can also be employed. Custom tuning can be performed based on at least query inputs in the form of text, graffiti, images, handwriting, voice, audio, and video signals. Natural language processing occurs, along with handwriting recognition and slang recognition. The system includes a classifier that receives a partial query as input, accesses a query database based on contents of the query input, and infers an intended search goal from query information stored on the query database. A query formulation engine receives search information associated with the intended search goal and generates a completed formal query for execution.

Claims

1. A computer-implemented system comprising:

one or more processors;

a recognition component that is executable by the one or more processors to receive a partial query input as voice signals of a user with at least one other sensed input type comprising text, image or audio;

a classifier component, executable by the one or more processors, that processes the partial query input and infers in real-time multiple different search goals based on the partial query input by accessing one or more query databases that store query information and from which similar or matching character sets, terms or phrases are derived for generating at least one complete query, the classifier comprising a support vector machine to find a hyperspace in a space of possible inputs to distinguish triggering input events from non-triggering input events, the classifier explicitly trained using generic training data and implicitly trained by observing user behavior;

a query formulation component, executable by the one or more processors, that generates the at least one complete query based on the multiple different search goals; and

a search engine, executable by the one or more processors, that receives the at least one complete query, presents the at least one complete query to the user for editing, and processes the at least one complete query to return search results for each of the at least one complete query, the search results based on a confidence value output by the voice recognition component indicating the confidence of converted voice signals relative to the partial query input as voice signals of the user.

2. The system of claim 1, wherein the search engine processes output results of the voice recognition component and preliminary search results to arrive at the search results.

3. The system of claim 1, wherein the search engine processes the multiple different search goals to return a set of search results for each different search goal.

4. The system of claim 1, further comprising a context component that generates context information related to a software environment in which the user is currently active, the context information employed by the classifier to infer the multiple different search goals.

5. The system of claim 1, further comprising a location component that computes geographic location data of the user and which location data the query formulation component processes to generate the at least one complete query.

6. The system of claim 1, wherein the classifier infers the multiple different search goals based on the partial query input as voice signals of the user and at least one other sensed input type.

7. The system of claim 1, further comprising a machine learning and reasoning component that learns and reasons about user interaction with the at least one complete query and ranks the search results based on the user interaction.

8. The system of claim 1, further comprising a syntactical processing component that receives recognized voice signals output from the voice recognition component and analyzes and processes word usage in the recognized voice signals for syntax properties.

9. The system of claim 1, further comprising a natural language processing component that receives recognized voice signals output from the voice recognition component and analyzes and processes word usage in the recognized voice signals for natural language characteristics.

10. A method comprising:

under control of one or more processors configured with executable instructions,

receiving a partial query input comprising voice signals of a user and at least one of text, image or audio input by a user;

inferring in real-time, using a classifier, multiple different search goals based on the partial query input, the classifier comprising a support vector machine to find a hyperspace in a space of possible inputs to distinguish triggering input events from non-triggering input events, the classifier explicitly trained using generic training data and implicitly trained by observing user behavior;

deriving similar or matching character sets, terms, or phrases to the partial query input by accessing one or more query databases that store query-related information;

formulating at least one complete query based on the multiple different search goals;

presenting the at least one complete query to the user for editing;

processing the at least one complete query to return search results;

assigning search results to each of the possible interpretations for selection by the user;

outputting a confidence value associated with each of the multiple different search goals;

formulating a complete query for each of the multiple different search goals;

executing each complete query to return search results; and

adjusting a number of the search results to be associated with each of the formal queries based at least in part on the associated confidence value.

11. The method of claim 10, further comprising:

outputting recognition results based on processing the partial query input; and

processing the recognition results in combination with preliminary search results obtained from processing the at least one complete query to arrive at a final set of search results.

12. The method of claim 10, further comprising processing historical information about the user and utilizing the processed historical information to generate the at least one complete query.

13. The method of claim 10, further comprising performing a search based on the at least one complete query, and storing ambiguities derived as part of the search for processing of a future search.

14. The method of claim 10, further comprising monitoring user interaction with the at least one complete query and, learning and reasoning about the user interaction to refine a future partial query input.

15. A computer-readable storage device comprising instructions executable by one or more processors to perform acts comprising:

receive a partial query including voice signals of a user and at least one of text, image or audio input by the user;

inferring in real-time, using a classifier, multiple different intended search goals based on the partial query input, the classifier comprising a support vector machine to find a hyperspace in a space of possible inputs to distinguish triggering input events from non-triggering input events, the classifier explicitly trained using generic training data and implicitly trained by observing user behavior;

deriving similar or matching character sets, terms, or phrases to the partial query input by accessing one or more query databases that store query-related information;

generating at least one complete query based on the multiple different search goals;

presenting the at least one complete query to the user for editing; and

processing the at least one complete query to return search results, the search results based on a confidence value indicating the confidence of converted voice signals relative to the partial query input as voice signals of the user.

16. The computer-readable storage device of claim 15, the acts further comprising tagging areas of a predefined ecosystem to enable processing the complete query based on the tagged areas.

17. The computer-readable storage device of claim 15, the acts further comprising formulating the complete query based on the inferred query information is operable to use alias processing to associate other characters, words, or phrases with the inferred query information to formulate the complete query.

18. The computer-readable storage device of claim 15, the acts further comprising learning user interaction information in response to the user interacting with the complete query, the learning based on edits or changes made by the user when interacting with the complete query.