Translating natural language utterances to keyword search queries

• US 10,061,843 B2
• Filed: 06/08/2015
• Issued: 08/28/2018
• Est. Priority Date: 05/12/2011
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Inventors: Hakkani-Tur, Dilek Zeynep, Tur, Gokhan, Iyer, Rukmini, Heck, Larry Paul
• Primary Examiner(s): LEROUX, ETIENNE PIERRE
• Application Number:
  US14/733,188
• Publication Number:
  US 20160004707A1
• Time in Patent Office: 1,177 Days
• CPC Class Codes: G06F 16/24522 Translation of natural language queries to structured queries
• CPC Class Codes: G06F 16/3329 Natural language query formulation or dialogue systems
• CPC Class Codes: G06F 16/3332 Query translation
• CPC Class Codes: G06F 16/3334 Selection or weighting of terms from queries, including natural language queries
• CPC Class Codes: G06F 16/3338 Query expansion
• CPC Class Codes: G06F 16/3344 using natural language analysis
• CPC Class Codes: G06F 16/84 Mapping; Conversion
• CPC Class Codes: G06F 16/951 Indexing; Web crawling techniques
• CPC Class Codes: G06F 40/44 Statistical methods, e.g. probability models
• CPC Class Codes: G06N 20/00 Machine learning

Abstract

Natural language query translation may be provided. A statistical model may be trained to detect domains according to a plurality of query click log data. Upon receiving a natural language query, the statistical model may be used to translate the natural language query into an action. The action may then be performed and at least one result associated with performing the action may be provided.

Claims

1. A method for providing natural language query translation, the method comprising:

training a statistical machine translation model according to a plurality of mined query pairs, wherein training the statistical machine translation model comprises;

identifying a set of previously received keyword type queries that are semantically similar to a set of previously received natural language type queries utilizing a query click graph,

pairing at least one keyword type query with a natural language type query that is most semantically similar to form the mined query pairs;

receiving a new natural language type query;

mapping the new natural language type query into a new keyword-based type query according to the trained statistical machine translation model;

performing a search query according to the new keyword-based type query; and

providing at least one result from the search query.

2. The method of claim 1, wherein the new natural language type query is received as text.

3. The method of claim 1, wherein the new natural language type query is received as speech.

4. The method of claim 1, wherein training the statistical machine translation model comprises identifying a plurality of domain independent salient phrases.

5. The method of claim 4, wherein at least one of the plurality of domain independent salient phrases comprises at least one word indicating that an associated search query comprises a natural language type search query.

6. The method of claim 4, wherein training the plurality of mined query pairs is associated with a plurality of search engine results.

7. The method of claim 6, further comprising identifying a plurality of search queries associated with the plurality of mined query pairs that comprise natural language type search queries according to the plurality of domain independent salient phrases.

8. The method of claim 1, wherein a correlation between a previously received natural language type query and a previously received keyword-based type query is associated with a Uniform Resource Locator (URL) distribution.

9. The method of claim 8, wherein performing the search query comprises:

searching the plurality of mined query pairs for a query pair corresponding to the search query; and

identifying a domain associated with the search query according to the URL distribution.

10. The method of claim 1, wherein the plurality of query pairs comprises a query pair associated with a geographic location.

11. The method of claim 1, wherein the at least one result is associated with at least one webpage.

12. A system for providing natural language query translation, the system comprising:

a memory storage; and

a processing unit coupled to the memory storage, wherein the processing unit is operable to;

train a statistical machine translation model according to a plurality of mined query pairs,

wherein train the statistical machine translation model comprises;

identifying a set of previously received keyword type queries that are semantically similar to a set of previously received natural language type queries utilizing a query click graph,

pairing at least one keyword type query with at least one natural language type query that is most semantically similar to form the mined query pair;

receive a query from a user,

determine whether the query is a natural language type query or is a keyword type query, and

in response to determining that the query comprises the natural language type query;

map the natural language type query into a keyword-based type query according to the trained statistical machine translation model;

perform a search query according to the natural language type query and the keyword-based type query; and

provide a plurality of results associated with the search query to the user.

13. The system of claim 12, wherein being operative to map the natural language type query into the keyword-based type query comprises being operative to:

detect a domain associated with the natural language type query; and

strip at least one domain-independent word from the natural language type query.

14. The system of claim 13, wherein being operative to detect the domain associated with the natural language type query comprises being operative to:

identify a subset of a plurality of possible domains according to at least one feature of the natural language type query.

15. The system of claim 14, wherein the at least one feature of the natural language type query comprises at least one of the following:

a lexical feature, a contextual feature, a semantic feature, a syntactic feature, and a topical feature.

16. The system of claim 14, wherein being operative to map the natural language type query into the keyword-based type query comprises being further operative to:

convert the natural language type query into the keyword-based type query according to the trained statistical machine translation model.

17. The system of claim 12, wherein a previously received natural language type query is identified according to a domain independent salient phrase.

18. The system of claim 12, wherein a previously received natural language type query and a previously received keyword-based type query are associated according to a weighted Uniform Resource Locator (URL) click graph.

19. The system of claim 12, wherein the mined query pairs comprises a statistical weighting according to a semantic correlation between a previously received natural language type query and a previously received keyword-based type query.

20. A mined query pair computer-readable medium which stores a set of instructions which when executed performs a method for providing natural language query translation, the method executed by the set of instructions comprising:

training a statistical machine translation model according to a plurality of mined query pairs,

wherein training the statistical machine translation model comprises;

identifying a plurality of domain independent salient phrases (DISPs),

identifying a plurality of previously received natural language type queries that correspond to the plurality of DISPs to form corresponding DISPs,

associating at least one of the plurality of previously received natural language type queries with a previously received keyword-based type query based on the corresponding DISPs to form a mined query pair of the plurality of mined query pairs according to a uniform resource locator (URL) click graph, wherein the URL click graph comprises a weighted distribution of URLs selected in response to the plurality of previous natural language queries and the previously received keyword-based type query, and

extracting a plurality of common features for at least one of the mined query pairs;

receiving a new search query from a user,

determining whether the new search query comprises a new natural language type search query,

in response to determining that the new search query comprises the new natural language type search query, mapping the new natural language type search query into a keyword-based type query according to the trained statistical machine translation model;

performing a search query according to the keyword-based type query; and

providing a plurality of results associated with the search query to the user.