Bootstrapping language models for spoken dialog systems using the world wide web

1. A method comprising:

identifying, via a processor communicating with Internet resources, common task independent web-sentences based on frequently occurring phrases across multiple websites from a web domain stored in a data store;

selectively removing the common task independent web-sentences from the web domain data, to yield filtered web domain data comprising domain-specific data;

identifying, via the processor, predicate/argument pairs from the filtered web domain data;

replacing, via the processor, the predicate/argument pairs with predicate/argument tokens;

generating, via the processor, conversational utterances by merging the predicate/argument tokens with manually written conversational templates while preserving a relative frequency of the manually written conversational templates, to yield generated conversational utterances; and

generating, via the processor, a web data language model using the generated conversational utterances, and providing it as an initial language model for deployment of an automated speech recognition system.

2. The method of claim 1 , wherein the predicates/argument pairs are identified by semantically parsing the filtered web domain data.

3. The method of claim 1 , wherein the predicate/argument tokens are used to identify specific tasks.

4. The method of claim 1 , wherein the conversational utterances are generated using manually written conversational templates according the relative frequency of occurrence.

5. The method of claim 1 , wherein the conversational utterances are generated by merging the predicate/argument tokens into the manually written conversational templates in combination with using a library of previously collected utterances.

6. The method of claim 1 , wherein the manually written conversational templates can be learned from a library of utterances from spoken dialogue systems.

7. A non-transitory computer-readable storage device having instructions stored which, when executed by a computing device connected to Internet resources, cause the computing device to perform operations-comprising:

identifying, via a processor communicating with Internet resources, common task independent web-sentences based on frequently occurring phrases across multiple websites from a web domain stored in a data store;

selectively removing the common task independent web-sentences from the web domain data, to yield filtered web domain data comprising domain-specific data;

identifying, via the processor, predicate/argument pairs from the filtered web domain data;

replacing, via the processor, the predicate/argument pairs with predicate/argument tokens;

generating, via the processor, conversational utterances by merging the predicate/argument tokens with manually written conversational templates while preserving a relative frequency of the manually written conversational templates, to yield generated conversational utterances; and

generating, via the processor, a web data language model using the generated conversational utterances, and providing it as an initial language model for deployment of an automated speech recognition system.

8. The non-transitory computer-readable storage device of claim 7 , wherein the predicates/argument pairs are identified by semantically parsing the filtered web domain data.

9. The non-transitory computer-readable storage device of claim 7 , wherein the predicate/argument tokens are used to identify specific tasks.

10. The non-transitory computer-readable storage device of claim 7 , wherein the conversational utterances are generated using the manually written conversational templates according the relative frequency of occurrence.

11. The non-transitory computer-readable storage device of claim 7 , wherein the conversational utterances are generated by merging the predicate/argument tokens into manually written conversational templates in combination with using a library of previously collected utterances.

12. The non-transitory computer-readable storage device of claim 7 , wherein the manually written conversational templates can be learned from a library of utterances from spoken dialogue systems.

13. A system comprising:

a processor connected to Internet resources; and

a computer readable storage medium having instructions stored which, when executed by the processor, cause the processor to perform operations comprising:

identifying, via the processor communicating with Internet resources, common task independent web-sentences based on frequently occurring phrases across multiple websites from a web domain stored in a data store;

selectively removing the common task independent web-sentences from the web domain data, to yield filtered web domain data comprising domain-specific data;

identifying, via the processor, predicate/argument pairs from the filtered web domain data;

replacing, via the processor, the predicate/argument pairs with predicate/argument tokens;

generating, via the processor, conversational utterances by merging the predicate/argument tokens with manually written conversational templates while preserving a relative frequency of the manually written conversational templates, to yield generated conversational utterances; and

generating, via the processor, a web data language model using the generated conversational utterances, and providing it as an initial language model for deployment of an automated speech recognition system.

14. The system of claim 13 , wherein the parser identifies predicates/argument pairs by semantically parsing the filtered web domain data.

15. The system of claim 13 , wherein the conversational utterances are generated using the manually written conversational templates according the relative frequency of occurrence.

16. The system of claim 13 , wherein the conversational utterances are generated by merging the predicate/argument tokens into manually written conversational templates in combination with using a library of previously collected utterances.

17. The system of claim 13 , wherein the manually written conversational templates can be learned from a library of utterances from spoken dialogue systems.