CLARIFYING NATURAL LANGUAGE INPUT USING TARGETED QUESTIONS

• US 20140316764A1
• Filed: 04/19/2013
• Published: 10/23/2014
• Est. Priority Date: 04/19/2013
• Status: Active Grant

• Current Assignee: SRI International, Inc.
• Original Assignee: SRI International, Inc.
• Inventors: Ayan, Necip Fazil, Zheng, Jing, Mandal, Arindam
• Granted Patent: US 9,805,718 B2
• Time in Patent Office: Days
• US Class Current: 704/9
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features
• CPC Class Codes: G10L 2015/225 Feedback of the input speech
• CPC Class Codes: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

A dialog assistant embodied in a computing system can present a clarification question based on a machine-readable version of human-generated conversational natural language input. Some versions of the dialog assistant identify a clarification target in the machine-readable version, determine a clarification type relating to the clarification target, present the clarification question in a conversational natural language manner, and process a human-generated conversational natural language response to the clarification question.

Claims

1. A dialog assistant embodied in one or more machine readable storage media, the dialog assistant accessible by a computing system to conduct a clarification dialog with a user by:

identifying a clarification target, the clarification target comprising a machine-readable version of a portion of a human-generated conversational natural language input needing clarification, the clarification target being less than the whole input;

determining a clarification type relating to the clarification target, the clarification type indicating a type of clarification needed to clarify the intended meaning of the clarification target;

presenting a conversational natural language clarification question based on the clarification type, the clarification question referring to the clarification target; and

processing a human-generated conversational natural language response to the clarification question to clarify the intended meaning of the clarification target.

2. The dialog assistant of claim 1, wherein the human-generated conversational natural language input comprises human-spoken conversational natural language input.

3. The dialog assistant of claim 1, wherein the human-generated conversational natural language response comprises human-spoken conversational natural language input.

4. The dialog assistant of claim 1, wherein the conversational natural language clarification question comprises machine-generated spoken conversational language output.

5. The dialog assistant of claim 1, wherein the computing system comprises a mobile electronic device.

6. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to present the clarification question at the computing system in a spoken natural language form.

7. The dialog assistant of claim 6, wherein the dialog assistant is accessible by the computing system to play at least a portion of the human-generated conversational natural language input through a speech synthesizer as part of the clarification question.

8. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to analyze the response and determine, based on the analysis of the response, whether to generate another clarification question.

9. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to extract an answer relating to the clarification target from the response and to modify the input by replacing at least a portion of the clarification target with a machine-readable version of at least a portion of the answer.

10. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate a new version of the input based on at least a portion of the clarification target.

11. The dialog assistant of claim 1, wherein the clarification target comprises at least one of:

a proper noun, a mispronunciation, a homophone, and an idiom.

12. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question by applying a clarification question template based on the clarification target and the clarification type.

13. The dialog assistant of claim 12, wherein the dialog assistant is accessible by the computing system to generate the clarification question to request that the response include the spelling of a word in the clarification target.

14. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question to request that the response include another word or phrase having a similar meaning to a word in the clarification target.

15. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question to present a plurality of homophones and request that the response indicate whether any of the homophones is the same as a word in the clarification target.

16. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question to present a word sense and request that the response indicate whether the word sense corresponds to the meaning of a word in the clarification target.

17. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question to include the input and replace a word of the clarification target with an interrogative pronoun.

18. The dialog assistant of claim 17, wherein the dialog assistant is accessible by the computing system to select the interrogative pronoun based on an attribute assigned to the clarification target by a software application.

19. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate the clarification question to include the clarification target and request that the response confirm the accuracy of the clarification target.

20. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to generate a plurality of clarification questions based on the clarification target, select one clarification question from the plurality of clarification questions, and present the selected clarification question at the computing system in a natural language form.

21. The dialog assistant of claim 20, wherein the dialog assistant is accessible by the computing system to analyze the response to the selected clarification question, select another clarification question from the plurality of clarification questions based on the analysis of the response, and present the other clarification question at the computing system in a natural language form.

22. The dialog assistant of claim 1, wherein the dialog assistant is accessible by the computing system to identify a plurality of different clarification targets in the input and generate a plurality of different clarification questions corresponding to the different clarification targets.

23. The dialog assistant of claim 22, wherein the dialog assistant is accessible by the computing system to process human-generated conversational natural language responses to each of the clarification questions to clarify the intended meaning of the clarification target.

24. The dialog assistant of claim 1, wherein the machine-readable version of the human-generated conversational natural language input comprises an output of one or more of an automatic speech recognition system and a statistical machine-translation system.

25. The dialog assistant of claim 24, wherein the machine-readable version of the human-generated conversational natural language input includes one or more attributes assigned to the input by the automatic speech recognition system or the statistical machine-translation system, and the dialog assistant is accessible by the computing system to generate the clarification question based on the one or more attributes assigned to the input.

26. The dialog assistant of claim 1, wherein the clarification type comprises one of a plurality of clarification types including an out-of-vocabulary phrase, an out-of-vocabulary word, an ambiguous phrase, and an ambiguous word and the clarification question comprises one of a plurality of clarification questions including clarification questions relating to each of the clarification types.

27. The dialog assistant of claim 1, wherein the dialog assistant is embodied in a virtual personal assistant.

28. The dialog assistant of claim 27, wherein the dialog assistant conducts the clarification dialog to clarify the input before the input is further processed by the virtual personal assistant.

29. The dialog assistant of claim 1, wherein the dialog assistant is embodied in a speech translation system.

30. The dialog assistant of claim 29, wherein the dialog assistant conducts the clarification dialog to clarify the input before the input is further processed by the speech translation system.

31. A method for conducting a clarification dialog with a user, the method comprising, with a computing system:

identifying a clarification target, the clarification target comprising a portion of a machine-readable version of a human-spoken conversational natural language input needing clarification, the clarification target being less than the whole input;

determining a clarification type relating to the clarification target, the clarification type indicating a type of clarification needed to clarify the intended meaning of the clarification target;

presenting a machine-generated spoken conversational natural language clarification question based on the clarification type, the clarification question referring to the clarification target; and

processing a human-spoken conversational natural language response to the clarification question to clarify the intended meaning of the clarification target.

32. The method of claim 31, wherein the processing comprises modifying the input to replace at least a portion of the clarification target with a machine-readable version of the response.

33. A computing device comprising:

a processor; and

a memory having stored therein a plurality of instructions that when executed by the processor cause the computing device to perform the method of claim 31.

34. One or more machine readable storage media comprising a plurality of instructions stored thereon that in response to being executed result in a computing device performing the method of claim 31.