Agent architecture for determining meanings of natural language utterances

• US 8,731,929 B2
• Filed: 02/04/2009
• Issued: 05/20/2014
• Est. Priority Date: 06/03/2002
• Status: Active Grant

• Current Assignee: Dialect, LLC
• Original Assignee: VoiceBox Technologies Corporation (Microsoft Corporation)
• Inventors: Kennewick, Robert A., Locke, David, Kennewick, Michael R. Sr., Kennewick, Michael R. Jr., Kennewick, Richard, Freeman, Tom
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US12/365,516
• Publication Number:
  US 20090171664A1
• Time in Patent Office: 1,931 Days
• Field of Search: 704/257, 704/270, 704/270.1, 704/275, 704/246, 707/709, 707/721, 707/751, 707/771
• US Class Current: 704/257
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/228 of application context
• CPC Class Codes: Y10S 707/99933 Query processing, i.e. searching

Abstract

Systems and methods for receiving natural language queries and/or commands and execute the queries and/or commands. The systems and methods overcomes the deficiencies of prior art speech query and response systems through the application of a complete speech-based information query, retrieval, presentation and command environment. This environment makes significant use of context, prior information, domain knowledge, and user specific profile data to achieve a natural environment for one or more users making queries or commands in multiple domains. Through this integrated approach, a complete speech-based natural language query and response environment can be created. The systems and methods creates, stores and uses extensive personal profile information for each user, thereby improving the reliability of determining the context and presenting the expected results for a particular question or command.

Claims

1. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

generate a digitized speech signal from the second natural language utterance;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

tag the one or more words in the second natural language utterance with a user identity determined from voice characteristics associated with the digitized speech signal and one or more user profiles;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request; and

receive a response to the request from the selected domain agent.

2. The system according to claim 1, wherein the computing device is further caused to recognize the one or more words in the second natural language utterance using information in the one or more dictionary and phrase tables, the one or more user profiles, and data supplied by one or more of the plurality of domain agents.

3. The system according to claim 1, wherein the computing device is further caused to:

create one or more response utterances to present a result generated from the selected domain agent processing the request; and

output the one or more response utterances via a speaker.

4. The system according to claim 3, wherein the selected domain agent is configured to select a presentation personality to format the result generated from the selected domain agent processing the request and use a template associated with the selected presentation personality to create a response string that includes the generated result, wherein the template is associated with a sarcastic personality, a humorous personality, a sympathetic personality, or an irritable personality.

5. The system according to claim 4, wherein the plurality of domain agents are each further configured to use the template associated with the selected presentation personality to format the response string based on the determined context, one or more criteria associated with the request, a dialog history, or a user profile.

6. The system according to claim 5, wherein the plurality of domain agents are each further configured to apply one or more variable substitutions or transformations to vary an order or a terminology used to present the generated results via the formatted response string.

7. The system according to claim 1, wherein the computing device is further caused to apply prior probabilities or fuzzy possibilities to one or more of keyword matching, a user profile, or a dialog history to determine the context.

8. The system according to claim 1, wherein to process the request, the plurality of domain agents are each configured to:

query one or more local or remote information sources in response to determining that the request includes the question.

9. The system according to claim 1, wherein to process the request, the plurality of domain agents are each configured to:

send multiple duplicate queries to multiple local or remote information sources in response to determining that the request includes the question; and

asynchronously evaluate responses associated with the multiple local or remote information sources processing the multiple duplicate queries.

10. The system according to claim 9, wherein to process the request, the plurality of domain agents are each further configured to use probabilistic or fuzzy scoring techniques to score the asynchronously evaluated responses based on one or more of previously received responses to the multiple duplicate queries, the determined context, one or more criteria associated with the question, a dialog history, a user profile, or information specific to one or more of the plurality of domain agents.

11. The system according to claim 1, wherein the computing device is further configured to:

receive the grammar from the selected domain agent; and

evaluate the determined context and the question using the grammar received from the selected domain agent, wherein the request includes all tokens that are required to format the question in the grammar used by the selected domain agent.

12. The system according to claim 11, wherein the request further includes one or more tokens that are optional to format the question in the grammar used by the selected domain agent.

13. The system according to claim 1, wherein the plurality of domain agents are part of an agent architecture that the computing device has access to, and wherein the agent architecture further includes a system agent configured to provide default functionality and services available to all of the plurality of domain agents, and wherein each of the plurality of domain agents uses a different grammar.

14. The system according to claim 13, wherein the system agent is further configured to manage one or more criteria handlers that the computing device is configured to use to determine the context, and wherein the one or more criteria handlers associated with the system agent are available to the system agent and all of the plurality of domain agents.

15. The system according to claim 13, wherein the agent architecture further includes an agent library that comprises one or more utilities available to the system agent and all of the plurality of domain agents to perform commonly used functions across all of the plurality of different domains.

16. The system according to claim 15, wherein the commonly used functions associated with the one or more utilities in the agent library include at least one of text and string handling, network communications, database lookup and management, fuzzy and probabilistic evaluation, or text to speech formats.

17. The system according to claim 13, wherein the agent architecture further includes an update manager configured to manage updates relating to one or more of the system agent, the plurality of domain agents, the agent library, one or more databases available to the agent architecture, or entries in one or more dictionary and phrase tables.

18. The system according to claim 17, wherein the updates include one or more of a new domain agent, additional domain knowledge associated with one or more of the plurality of domain agents, new keywords associated with one or more of the plurality of domain agents, preferred information sources associated with one or more of the plurality of domain agents, updated domain information associated with one or more of the plurality of domain agents, or updated content associated with one or more of the plurality of domain agents.

19. The system according to claim 18, wherein the agent manager is further configured to:

use a network interface to locate the new domain agent in response to determining that none of the plurality of domain agents currently loaded in the agent architecture are suitable to process the request; and

cause the update manager to load the new domain agent located via the network interface pursuant to the terms and conditions of the license and invoke the new domain agent to process the request.

20. The system according to claim 17, wherein the update manager is configured uninstall one or more of the plurality of domain agents that are unused pursuant to a license with a third party to manage the updates relating to the plurality of domain agents.

21. The system according to claim 20, wherein the agent architecture further includes an agent manager configured to verify compliance with terms and conditions of the license.

22. The system according to claim 13, wherein the agent architecture further includes an agent manager configured to:

load and initialize the system agent and the plurality of domain agents when the computing device starts up; and

unload the system agent and the plurality of domain agents when the computing device shuts down.

23. The system according to claim 13, wherein the computing device is further configured to recognize the one or more words in the second natural language utterance using information in the one or more dictionary and phrase tables, one or more user profiles, and data supplied by one or more of the plurality of domain agents that are associated with the one or more of the plurality of different domains.

24. The system according to claim 1, further comprising an extension and modification facility configured to:

enable behavior associated with the agent architecture to be configured via one or more speech interface commands or one or more inputs to a graphical user interface; and

enable behavior associated with one or more of the plurality of domain agents to be configured via the one or more speech interface commands or the one or more inputs to the graphical user interface.

25. The system according to claim 24, wherein the one or more speech interface commands or the one or more inputs to the graphical user interface can add one or more of new commands, keywords, information, or information sources to configure the behavior associated with the agent architecture or the behavior associated with one or more of the plurality of domain agents.

26. The system according to claim 24, wherein the extension and modification facility includes a generic agent that the one or more speech interface commands or the one or more inputs to the graphical user interface can use to create a new domain agent to add to the agent architecture and configure the behavior associated therewith.

27. The system according to claim 24, wherein the one or more speech interface commands or the one or more inputs to the graphical user interface can use a scripting language to configure the behavior associated with the agent architecture or the behavior associated with the one or more of the plurality of domain agents.

28. The system according to claim 24, wherein the one or more speech interface commands or the one or more inputs to the graphical user interface can create one or more of special reports, query packages, alerts, or output formats to configure the behavior associated with the agent architecture or the behavior associated with the one or more of the plurality of domain agents.

29. The system according to claim 1, wherein the computing device comprises a parser that is configured to determine the context of the question.

30. The system according to claim 29, wherein the parser is configured to determine the context of the question from a dialog history that comprises previous natural language utterances received by the computing device.

31. The system according to claim 30, wherein the previous natural language utterances of the dialog history are all related to a thread of inquiry.

32. The system according to claim 1, wherein the computing device is further caused to determine the context by assigning a score to each of a plurality of candidate contexts based on age of the candidate context.

33. The system according to claim 32, wherein the score for each of the plurality of candidate contexts is further based on a user profile and on data supplied by one or more of the plurality of domain agents.

34. The system according to claim 1, wherein the question is a first question, and wherein the computing device is further caused to:

identify one or more parameters of the first question based on one or more words of the natural language utterance;

receive a subsequent natural language utterance that comprises a second question, wherein the second question is related to the first question; and

generate a second request based on the subsequent natural language utterance, the one or more parameters, and the grammar, wherein the second request includes the second question.

35. The system of claim 34, wherein the second question is a follow-up question of the first question.

36. The system of claim 35, wherein the computing device is further caused to store a value of the one or more parameters in a context stack and to transform the subsequent natural language utterance into the second request by retrieving the value of the one or more parameters from the context stack.

37. A system for processing natural language utterances, comprising:

one or more physical processors programmed to execute one or more computer program instructions which, when executed, cause the one or more physical processors to;

receive a natural language utterance;

determine that one or more words of the natural language utterance were unrecognized or incorrectly recognized in response to a recognition associated with the natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

identify, in one or more dictionary and phrase tables, one or more words that correspond to the one or more unrecognized or incorrectly recognized words based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables with respect to the one or more corresponding words based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a subsequent natural language utterance;

generate a digitized speech signal from the subsequent natural language utterance;

recognize one or more words in the subsequent natural language utterance based on a pronunciation associated with the one or more words in the subsequent natural language utterance using the one or more updated dictionary and phrase tables; and

tag the one or more words in the subsequent natural language utterance with a user identity determined from voice characteristics associated with the digitized speech signal and one or more user profiles.

38. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a command;

generate a digitized speech signal from the second natural language utterance;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

tag the one or more words in the second natural language utterance with a user identity determined from voice characteristics associated with the digitized speech signal and one or more user profiles;

determine a context of the command in the second natural language utterance;

select one of the plurality of domain agents based on the context of the command;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the command; and

invoke the selected domain agent to cause the selected domain agent to process the request.

39. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request;

generate one or more response utterances to present a result generated from the selected domain agent, wherein the selected domain agent is configured to select a presentation personality to format the result and use a template associated with the selected presentation personality to generate a response string that includes the generated result, wherein the template is associated with a sarcastic personality, a humorous personality, a sympathetic personality, or an irritable personality; and

output the one or more response utterances via a speaker.

40. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request, wherein to process the request, the plurality of domain agents are each configured to send multiple duplicate queries to multiple local or remote information sources in response to determining that the request includes the question, and asynchronously evaluate responses associated with the multiple local or remote information sources; and

receive a response to the request from the selected domain agent.

41. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request; and

receive a response to the request from the selected domain agent, wherein the plurality of domain agents are part of an agent architecture that the computing device has access to, wherein the agent architecture further includes a system agent configured to (i) provide default functionality and services available to the plurality of domain agents, (ii) manage one or more criteria handlers that the computing device is configured to use to determine the context, wherein the one or more criteria handlers associated with the system agent are available to the system agent and the plurality of domain agents, and wherein the plurality of domain agents use different grammars.

42. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

receive a grammar from the selected domain agent;

evaluate the determined context and the question using the grammar, wherein the request includes all tokens that are required to format the question based on the grammar;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question and one or more tokens that are optional to format the question based on the grammar;

invoke the selected domain agent to cause the selected domain agent to process the request; and

receive a response to the request from the selected domain agent.

43. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request; and

receive a response to the request from the selected domain agent, wherein the plurality of domain agents are part of an agent architecture that the computing device has access to, wherein the agent architecture further includes (i) a system agent configured to provide default functionality and services available to the plurality of domain agents, and (ii) an update manager configured to (a) manage updates relating to one or more of the system agent, the plurality of domain agents, the agent library, one or more databases available to the agent architecture, or entries in the one or more dictionary and phrase tables, and (b) uninstall one or more of the plurality of domain agents that are unused pursuant to a license with a third party to manage the updates relating to the plurality of domain agents, and wherein the plurality of domain agents use different grammars.

44. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request; and

receive a response to the request from the selected domain agent, wherein the plurality of domain agents are part of an agent architecture that the computing device has access to, wherein the agent architecture further includes (i) a system agent configured to provide default functionality and services available to the plurality of domain agents, and (ii) an update manager configured to manage updates relating to one or more of the system agent, the plurality of domain agents, the agent library, one or more databases available to the agent architecture, or entries in the one or more dictionary and phrase tables, wherein the updates include one or more of a new domain agent, additional domain knowledge associated with one or more of the plurality of domain agents, new keywords associated with one or more of the plurality of domain agents, preferred information sources associated with one or more of the plurality of domain agents, updated domain information associated with one or more of the plurality of domain agents, or updated content associated with one or more of the plurality of domain agents, and wherein the plurality of domain agents use different grammars, wherein the system agent is further configured to (i) use a network interface to locate the new domain agent in response to determining that none of the plurality of domain agents currently loaded in the agent architecture are suitable to process the request, and (ii) cause the update manager to load the new domain agent located via the network interface pursuant to the terms and conditions of the license and invoke the new domain agent to process the request.

45. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a question;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the question in the second natural language utterance by assigning a score to each of a plurality of candidate contexts based on age of the candidate context;

select one of the plurality of domain agents based on the context of the question;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request, wherein to process the request, the plurality of domain agents are each configured to send multiple duplicate queries to multiple local or remote information sources in response to determining that the request includes the question, and asynchronously evaluate responses associated with the multiple local or remote information sources; and

receive a response to the request from the selected domain agent.

46. A system for processing natural language utterances, comprising:

one or more physical processors programmed to execute one or more computer program instructions which, when executed, cause the one or more physical processors to;

receive a natural language utterance;

determine that one or more words of the natural language utterance were unrecognized or incorrectly recognized in response to a recognition associated with the natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

identify, in one or more dictionary and phrase tables, one or more words that correspond to the one or more unrecognized or incorrectly recognized words based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables with respect to the one or more corresponding words based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a subsequent natural language utterance that comprises a question;

recognize one or more words in the subsequent natural language utterance based on a pronunciation associated with the one or more words in the subsequent natural language utterance using the one or more updated dictionary and phrase tables;

determine a context of the question in the subsequent natural language utterance;

select one of a plurality of domain agents based on the context;

generate a request associated with the subsequent natural language utterance based on the one or more words in the subsequent natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question;

invoke the selected domain agent to cause the selected domain agent to process the request;

generate one or more response utterances to present a result generated from the selected domain agent, wherein the selected domain agent is configured to select a presentation personality to format the result and use a template associated with the selected presentation personality to generate a response string that includes the generated result, wherein the template is associated with a sarcastic personality, a humorous personality, a sympathetic personality, or an irritable personality; and

output the one or more response utterances via a speaker.

47. A system for processing natural language utterances, comprising:

one or more physical processors programmed to execute one or more computer program instructions which, when executed, cause the one or more physical processors to;

receive a natural language utterance;

determine that one or more words of the natural language utterance were unrecognized or incorrectly recognized in response to a recognition associated with the natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

identify, in one or more dictionary and phrase tables, one or more words that correspond to the one or more unrecognized or incorrectly recognized words based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables with respect to the one or more corresponding words based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a subsequent natural language utterance that comprises a question;

recognize one or more words in the subsequent natural language utterance based on a pronunciation associated with the one or more words in the subsequent natural language utterance using the one or more updated dictionary and phrase tables;

determine a context of the question in the subsequent natural language utterance;

select one of a plurality of domain agents based on the context;

generate a request associated with the subsequent natural language utterance based on the one or more words in the subsequent natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question; and

invoke the selected domain agent to cause the selected domain agent to process the request, wherein to process the request, the plurality of domain agents are each configured to send multiple duplicate queries to multiple local or remote information sources in response to determining that the request includes the question, and asynchronously evaluate responses associated with the multiple local or remote information sources.

48. A system for processing natural language utterances, comprising:

one or more physical processors programmed to execute one or more computer program instructions which, when executed, cause the one or more physical processors to;

receive a natural language utterance;

determine that one or more words of the natural language utterance were unrecognized or incorrectly recognized in response to a recognition associated with the natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

identify, in one or more dictionary and phrase tables, one or more words that correspond to the one or more unrecognized or incorrectly recognized words based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables with respect to the one or more corresponding words based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a subsequent natural language utterance that comprises a question;

recognize one or more words in the subsequent natural language utterance based on a pronunciation associated with the one or more words in the subsequent natural language utterance using the one or more updated dictionary and phrase tables;

determine a context of the question in the subsequent natural language utterance;

select one of a plurality of domain agents based on the context;

receive a grammar from the selected domain agent;

evaluate the determined context and the question using the grammar, wherein the request includes all tokens that are required to format the question based on the grammar;

generate a request associated with the subsequent natural language utterance based on the one or more words in the subsequent natural language utterance and the grammar, wherein the request includes the question and one or more tokens that are optional to format the question based on the grammar; and

invoke the selected domain agent to cause the selected domain agent to process the request.

49. A system for processing natural language utterances, comprising:

one or more physical processors programmed to execute one or more computer program instructions which, when executed, cause the one or more physical processors to;

receive a natural language utterance;

determine one or more words of the natural language utterance were unrecognized or incorrectly recognized in response to a recognition associated with the natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

identify, in one or more dictionary and phrase tables, one or more words that correspond to the one or more unrecognized or incorrectly recognized words based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables with respect to the one or more corresponding words based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a subsequent natural language utterance that comprises a question;

recognize one or more words in the subsequent natural language utterance based on a pronunciation associated with the one or more words in the subsequent natural language utterance using the one or more updated dictionary and phrase tables;

determine a context of the question in the subsequent natural language utterance by assigning a score to each of a plurality of candidate contexts based on age of the candidate context;

select one of a plurality of domain agents based on the context;

generate a request associated with the subsequent natural language utterance based on the one or more words in the subsequent natural language utterance and a grammar used by the selected domain agent, wherein the request includes the question; and

invoke the selected domain agent to cause the selected domain agent to process the request, wherein to process the request, the plurality of domain agents are each configured to send multiple duplicate queries to multiple local or remote information sources in response to determining that the request includes the question, and asynchronously evaluate responses associated with the multiple local or remote information sources.

50. A system for processing natural language utterances, comprising:

a computing device having access to a plurality of domain agents associated with a plurality of different domains, and programmed to execute one or more computer program instructions which, when executed, cause the computing device to;

receive a first natural language utterance;

determine that the first natural language utterance contains one or more words that were unrecognized or incorrectly recognized in response to a recognition associated with the first natural language utterance having a confidence level below a predetermined value;

obtain a phonetic alphabet spelling associated with the one or more unrecognized or incorrectly recognized words in response to the determination;

look up the one or more unrecognized or incorrectly recognized words in one or more dictionary and phrase tables based on the phonetic alphabet spelling;

update the one or more dictionary and phrase tables based on a pronunciation associated with the one or more unrecognized or incorrectly recognized words;

receive a second natural language utterance that comprises a command;

recognize one or more words in the second natural language utterance based on a pronunciation associated with the one or more words using the one or more dictionary and phrase tables;

determine a context of the command in the second natural language utterance;

select one of the plurality of domain agents based on the context of the command;

receive a grammar from the selected domain agent;

evaluate the determined context and the command using the grammar, wherein the request includes all tokens that are required to format the command based on the grammar;

generate a request associated with the second natural language utterance based on the one or more words in the second natural language utterance and a grammar used by the selected domain agent, wherein the request includes the command and one or more tokens that are optional to format the command based on the grammar; and

invoke the selected domain agent to cause the selected domain agent to process the request.