Relevance recognition for a human machine dialog system contextual question answering based on a normalization of the length of the user input

• US 8,204,751 B1
• Filed: 07/24/2006
• Issued: 06/19/2012
• Est. Priority Date: 03/03/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property II LP (AT&T, Inc.)
• Inventors: Di Fabbrizio, Giuseppe, Feng, Junlan
• Primary Examiner(s): Smits, Talivaldis Ivars
• Assistant Examiner(s): Kazeminezhad, Farzad
• Application Number:
  US11/459,523
• Time in Patent Office: 2,157 Days
• Field of Search: 704/275, 704/277, 704/256
• US Class Current: 704/275
• CPC Class Codes: G06F 40/211 Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G06F 40/58 Use of machine translation, e.g. for multi-lingual retrieval, for server-side translation for client devices or for real-time translation
• CPC Class Codes: G10L 15/142 Hidden Markov Models [HMMs]
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

Disclosed are systems, methods and computer-readable media for controlling a computing device to provide contextual responses to user inputs. The method comprises receiving a user input, generating a set of features characterizing an association between the user input and a conversation context based on at least a semantic and syntactic analysis of user inputs and system responses, determining with a data-driven machine learning approach whether the user input begins a new topic or is associated with a previous conversation context and if the received question is associated with the existing topic, then generating a response to the user input using information associated with the user input and any previous user input associated with the existing topic.

Claims

1. A method of automatically providing interactive question answering in a continuous dialog with a user, the method comprising:

receiving a current user input as part of the continuous dialog;

generating a set of features characterizing an association between the current user input and a conversation context based on at least a semantic and syntactic analysis of prior user inputs received before the current user input and system natural language responses to the prior user inputs within the continuous dialog, wherein a same user provides the current user input and the prior user inputs;

determining, based on a normalization of a length of the user input and a previous user input of the prior user inputs, with a data-driven machine learning approach whether the current user input begins a new topic of the continuous dialog or is associated with an existing topic related to a previous conversation context within the continuous dialog;

and if the current user input is associated with the existing topic, then generating a response to the current user input using information associated with the current user input and content in any previous user input associated with the existing topic.

2. The method of claim 1, wherein generating response to the current user input further comprises using a model trained for context fusion.

3. The method of claim 1, wherein the previous conversation context is associated with a topic.

4. The method of claim 1, wherein the data-driven machine learning approach is applied using one of a decision tree, Adaboost, SVM or MaxEnt.

5. The method of claim 1, wherein a user input is a natural language request.

6. The method of claim 1, wherein the training data used to learn relevancy recognition and contextual information fusion models for interactive question answering comprises at least one of:

interactive conversations, a series of questions on a same topic, or a sequence of question/answer pairs on related topics.

7. The method of claim 1, wherein the prior user inputs are classified based on context space.

8. The method of claim 1, wherein generating a response to the current user input using information associated with the current user input and content in any previous user input associated with the existing topic further comprises using an information fusion of the current user input and the previous user input.

9. The method of claim 1, wherein the prior user inputs are classified based on context space.

10. A system for providing interactive question answering in a continuous dialog with a user, the method comprising:

a first module configured to control a processor to receive a current user input as part of the continuous dialog;

a second module configured to control the processor to generate a set of features characterizing an association between the current user input and a conversation context based on at least a semantic and syntactic analysis of prior user inputs received before the current user input and system natural language responses to the prior user inputs within the continuous dialog, wherein a same user provides the current user input and the prior user inputs;

a third module configured to control the processor to determine, based on a normalization of a length of the user input and a previous user inputs, with a data-driven machine learning approach whether the current user input is one of beginning a new topic within the continuous dialog or is associated with an existing topic related to a previous conversation context of the continuous dialog; and

a fourth module configured, if the current user input is associated with the existing topic, to control the processor to generate a response to the current user input using information associated with the current user input and content in any prior user inputs associated with the existing topic.

11. The system of claim 10, wherein the fourth module further uses a model trained for context fusion.

12. The system of claim 10, wherein the previous conversation context is associated with a topic.

13. The system of claim 10, wherein the data-driven machine learning approach is applied using one of a decision tree, Adaboost, SVM or Maxent.

14. The system of claim 10, wherein a user input is a natural language request.

15. The system of claim 10, wherein the prior user inputs are classified based on context space.

16. The system of claim 10, wherein generating a response to the current user input using information associated with the current user input and content in any prior user inputs associated with the existing topic further comprises using an information fusion of the current user input and the prior user inputs.

17. A non-transitory computer-readable storage medium storing instructions for providing interactive question answering in a continuous dialog with a user, the instructions comprising computer-executable instructions for:

receiving a current user input as part of the continuous dialog;

generating a set of features characterizing an association between the current user input and a conversation context based on at least a semantic and syntactic analysis of prior user inputs received before the current user input and system natural language responses to the prior user inputs within the continuous dialog, wherein a same user provides the current user input and the prior user inputs;

determining, based on a normalization of a length of the user input and a previous user input of the prior user inputs, with a data-driven machine learning approach whether the current user input is one of beginning a new topic within the continuous dialog or is associated with an existing topic related to a previous conversation context of the continuous dialog; and

if the current user input is associated with the existing topic, then generating a response to the current user input using information associated with the current user input and content in any previous user input of the prior user inputs associated with the existing topic.

18. A non-transitory computer-readable storage medium of claim 17, wherein generating a response to the current user input further comprises using a model trained for context fusion.

19. A non-transitory computer-readable storage medium of claim 18, wherein training data used to learn relevancy recognition and contextual information fusion models for interactive question answering comprises at least one of:

interactive conversations, a series of questions on a same topic, or a sequence of question/answer pairs on related topics.