Method and system for predicting understanding errors in a task classification system

• US 7,487,088 B1
• Filed: 05/12/2006
• Issued: 02/03/2009
• Est. Priority Date: 11/15/2000
• Status: Expired due to Term

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Geary, Irene Langkilde, Wright, Jeremy H., Gorin, Allen Louis, Walker, Marilyn Ann
• Primary Examiner(s): Lerner, Martin
• Application Number:
  US11/433,392
• Time in Patent Office: 998 Days
• Field of Search: 704/240, 704/252, 704/257, 704/270, 704/275, 704/231, 704/243, 704/255
• US Class Current: 704/240
• 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: H04M 3/4936 Speech interaction details speech recognition per se G10L15/00

Abstract

This invention concerns a method and system for monitoring an automated dialog system for the automatic recognition of language understanding errors based on a user'"'"'s input communications. The method may include determining whether a probability of understanding the user'"'"'s input communication exceeds a first threshold. If the first threshold is exceeded, further dialog is conducted with the user. Otherwise, the user may be directed to a human for assistance. The method also illustratively determines whether the probability also exceeds a second threshold, the second threshold being higher than the first. If so, then further dialog is conducted with the user using the current dialog strategy. However, if the probability falls between a first threshold and a second threshold, the dialog strategy may be adapted in order to improve the chances of conducting a successful dialog with the user.

Claims

1. A method of monitoring natural language understanding of a user'"'"'s input communication in a task classification system that operates on a task objective of a user, the method comprising

a) if the user'"'"'s input communication is sufficiently free of natural language understanding (NLU) errors that the user'"'"'s input communication can be understood, then making a task classification decision, and

b) if the user'"'"'s input communication contains NLU errors that make the user'"'"'s input communication not able to be understood, then

i) if a probability of understanding the user'"'"'s input communication exceeds a first threshold, conducting further dialog with the user,

ii) otherwise, routing the user to a human for further assistance.

2. The method of claim 1 further comprising

if a probability of understanding the user'"'"'s input communication exceeds a second threshold, the second threshold being greater than the first threshold, then conducting further dialog with the user using a current dialog strategy.

3. The method of claim 2 wherein if the second threshold is not exceeded, conducting further dialog with the user using an adapted dialog strategy.

4. The method of claim 3 wherein the adapted dialog strategy includes one of prompting the user with choices and prompting the user to confirm the recognition of NLU errors.

5. The method of claim 1 wherein the probability is determined using recognition of NLU errors derived from the user'"'"'s input communication.

6. The method of claim 1 wherein the probability is determined using training data stored in a training database, the training data including at least one of classification models and extracted features.

7. The method of claim 6 wherein the extracted features are derived from recognition, understanding and dialog data.

8. The method of claim 7, further comprising

storing a first dialog exchange in a dialog history database, wherein the first dialog exchange includes a first automated dialog output and the user'"'"'s first input communication and the further dialog conducted with the user includes a second dialog exchange which includes a second dialog output and the user'"'"'s second input communication;

and

determining whether the probability of understanding exceeds the first threshold using the first dialog exchange and the second dialog exchange.

9. The method of claim 8 wherein the method is recursive in that the determining step determines whether the probability of understanding exceeds the first threshold using each of the dialog exchanges conducted.

10. The method of claim 1 wherein the task classification decision includes classifying the task as one of a set of task objectives, and wherein if a task classification decision is made, the user'"'"'s input communication is routed in accordance with the classified task objective.

11. The method of claim 1 further comprising

recognizing portions of the user'"'"'s input communication;

and

providing an input to a language understanding monitor based on applying a confidence function to the recognized portions of the user'"'"'s input communication.