Method and apparatus for converting utterance representations into actions in a conversational system

• US 7,127,402 B2
• Filed: 01/10/2002
• Issued: 10/24/2006
• Est. Priority Date: 01/12/2001
• Status: Expired due to Term

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: International Business Machines Corporation
• Inventors: Ross, Steven I., Brownholtz, Elizabeth A., MacAllister, Jeffrey G.
• Primary Examiner(s): HARPER, V PAUL
• Application Number:
  US10/044,760
• Publication Number:
  US 20020138266A1
• Time in Patent Office: 1,748 Days
• Field of Search: 704/1, 704/231, 704/270, 704/275
• US Class Current: 704/275
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 2015/228 of application context

Abstract

A conversation manager processes a spoken utterance from a user of a computer that is directed to an application program hosted on the computer. The conversation manager includes a reasoning facility which accesses goal-directed rules stored in a rules base (e.g., database). The reasoning facility also has access to a conversational record that includes a record of previous utterances and a semantic analysis for each utterance. The reasoning facility processes a representation of the utterance by using the goal-directed rules. The reasoning facility uses means-end analysis to determine the proper rules to execute, and thus the script calls to make to achieve the goal of processing the utterance. While processing the utterance, the reasoning facility attempts to resolve any ambiguities in the representation of the utterance and to fill in any missing information that is needed to achieve its goal. The reasoning facility then generates a response to the utterance, which can be a question to the user or a command to the application program based on the utterance.

Claims

1. A computer method for converting an utterance representation into a response, the computer method comprising the steps of:

generating an application specific goal derived from the utterance representation, wherein the application specific goal and the utterance representation are propositions comprising attribute-object-value triples, the proposition corresponding to the utterance representation being derived from a frame representation;

analyzing the utterance representation based on the application specific goal and a corresponding set of goal-directed rules to identify ambiguous information in the utterance representation or to generate a subgoal, each goal-directed rule representing a valid inference step from the goal-directed rule'"'"'s condition to the application specific goal that may be taken in a domain associated with the application specific goal;

generating a response based on the analysis of the utterance representation if ambiguous information is identified; and

wherein the response is a computer application program command based on the utterance representation.

2. The computer method of claim 1 wherein the step of analyzing the utterance representation comprises applying a goal-directed reasoning analysis based on the set of goal-directed rules to clarify the ambiguous information.

3. The computer method of claim 2, wherein the step of analyzing the utterance representation comprises accessing data in a conversational record of related utterances to clarify the ambiguous information.

4. The computer method of claim 2, wherein the step of generating the response comprises generating a question directed to a provider of the utterance representation to clarify the ambiguous information, the question emerging from the analyzing of the utterance representation and requesting further information from the provider.

5. The computer method of claim 1, wherein the step of generating the response comprises generating the computer application program command based on the utterance representation and based on the analysis of the ambiguous information.

6. The computer method of claim 1, wherein each goal-directed rule comprises a set of conditions and a set of actions, each condition consisting of a first proposition or a first script command and each action consisting of a second proposition or a second script command.

7. An apparatus for converting an utterance representation into a response, comprising:

a database storing a set of goal-directed rules;

a digital processor coupled to the database, the digital processor hosting and executing a reasoning facility that is configured to;

generate an application specific goal derived from the utterance representation, wherein the application specific goal and the utterance representation are propositions comprising attribute-object-value triples, the proposition corresponding to the utterance representation being derived from a frame representation;

analyze the utterance representation based on the application specific goal and the set of goal-directed rules in the database to identify ambiguous information in the utterance representation or to generate a subgoal, each goal-directed rule representing a valid inference step from the goal-directed rule'"'"'s condition to the application specific goal that may be taken in a domain associated with the application specific goal; and

generate a response based on the analysis of the utterance representation if ambiguous information is identified.

8. The apparatus of claim 7 wherein the reasoning facility applies a goal-directed reasoning analysis based on the set of goal-directed rules to clarify the ambiguous information.

9. The apparatus of claim 8, wherein the reasoning facility accesses data in a conversational record of related utterances to clarify the ambiguous information.

10. The apparatus of claim 8, wherein the reasoning facility generates a question directed to a provider of the utterance representation to clarify the ambiguous information, the question emerging from the analysis of the utterance representation and requesting further information from the provider.

11. The apparatus of claim 7, wherein the reasoning facility generates the computer application program command based on the utterance representation and based on the analysis of the ambiguous information.

12. The apparatus of claim 7, wherein each goal-directed rule comprises a set of conditions and a set of actions, each condition consisting of a first proposition or a first script command and each action consisting of a second proposition or a second script command.

13. The apparatus of claim 7, wherein the response is a computer application program command based on the utterance representation.

14. A computer program product comprising:

a computer for converting an utterance representation into a response; and

a set of computer program instructions embodied on the computer readable medium, including instructions to;

generate an application specific goal derived from the utterance representation, wherein the application specific goal and the utterance representation are propositions comprising attribute-object-value triples, the proposition corresponding to the utterance representation being derived from a frame representation;

analyze the utterance representation based on the application specific goal and a corresponding set of goal-directed rules to identify ambiguous information in the utterance representation or to generate a subgoal, each goal-directed rule representing a valid inference step from the goal-directed rule'"'"'s condition to the application specific goal that may be taken in a domain associated with the application specific goal;

generate a response based on the analysis of the utterance representation if ambiguous information is identified; and

wherein the response is a computer application program command based on the utterance representation.

15. The computer program product of claim 14 wherein the set of computer instructions comprises further instructions to apply a goal-directed reasoning analysis based on the set of goal-directed rules to clarify the ambiguous information.

16. The computer program product of claim 15, wherein the set of computer instructions comprises further instructions to access data in a conversational record of related utterances to clarify the ambiguous information.

17. The computer program product of claim 15, wherein the set of computer instructions comprises further instructions to generate a question directed to a provider of the utterance representation to clarify the ambiguous information, the question emerging from the analysis of the utterance representation and requesting further information from the provider.

18. The computer program product of claim 14, wherein the set of computer instructions comprises further instructions to generate the computer application program command based on the utterance representation and based on the analysis of the ambiguous information.

19. The computer program product of claim 14, wherein each goal-directed rule comprises a set of conditions and a set of actions, each condition consisting of a first proposition or a first script command and each action consisting of a second proposition or a second script command.

20. An apparatus for converting an utterance representation into a response, comprising:

means for generating an application specific goal derived from the utterance representation, wherein the application specific goal and the utterance representation are propositions comprising attribute-object-value triples, the proposition corresponding to the utterance representation being derived from a frame representation;

means for analyzing the utterance representation based on the application specific goal and a corresponding set of goal-directed rules to identify ambiguous information in the utterance representation or to generate a subgoal, each goal-directed rule representing a valid inference step from the goal-directed rule'"'"'s condition to the application specific goal that may be taken in a domain associated with the application specific goal; and

means for generating a response based on the analysis of the utterance representation if ambiguous information is identified.