System and method for context sensitive inference in a speech processing system

Patent metadata

Field	Value
Publication number	US-9626968-B2
Application number	US-201514594570-A
Country	US
Kind code	B2
Filing date	Jan 12, 2015
Priority date	Jun 25, 2008
Publication date	Apr 18, 2017
Grant date	Apr 18, 2017

Abstract

A method of operating a speech processing system is provided. The method includes translating a portion of a speech record into a plurality of possible words associated with a plurality of contexts, and determining a plurality of correctness values based on a plurality of probabilities that each of the plurality of possible words is correct for each of the plurality of contexts. The method also includes determining which of the plurality of possible words is a correct translation of the portion of the speech record based on the plurality of correctness values.

First claim

What is claimed is:

1. A method of operating a speech processing system, the method comprising:
receiving a portion of a speech record from an audio source, the speech record having been produced by the audio source and transferred by the audio source to the speech processing system;
determining, by the speech processing system, from metadata accompanying the portion of the speech record, a first context and a second context for the portion of the speech record, wherein the first context has a first probability of correct context and the second context has a second probability of correct context;
processing, by the speech processing system, the portion of the speech record to create a first text translation for the portion of the speech record in the first context, wherein the first text translation has a first probability of correct translation within the first context;
processing, by the speech processing system, the same portion of the speech record to create a second text translation for the portion of the speech record in the second context, wherein the second text translation has a second probability of correct translation within the second context;
processing, by the speech processing system, the first probability of correct translation and the first probability of correct context, to produce a first probability;
processing, by the speech processing system, the second probability of correct translation within the second context and the second probability of correct context, to produce a second probability;
selecting, by the speech processing system, the first translation as the correct translation when the first probability is greater than the second probability; and
selecting, by the speech processing system, the second translation as the correct translation when the second probability is greater than the first probability.

2. The method of claim 1 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is an identity of a speaker.

3. The method of claim 1 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is a location of a speaker.

4. The method of claim 1 , wherein the portion of the speech record is a portion of a sentence and the context for the portion of the speech record is a position of the portion of the speech record within the sentence.

5. The method of claim 1 , wherein the portion of the speech record is a portion of a call and the context for the portion of the speech record is a position of the portion of the speech record within the call.

6. The method of claim 1 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is a position of the portion of the speech record within the conversation.

7. The method of claim 1 , wherein the first probability of correct translation within the first context is multiplied by the first probability of correct context, resulting in the first probability.

8. A non-transitory computer readable medium having stored thereon instructions that, when executed by processing circuitry, direct the processing circuitry to perform the steps comprising:
receiving a portion of a speech record from an audio source, the speech record having been produced by the audio source and transferred to the processing circuitry;
determining, from metadata accompanying the portion of the speech record, a first context and a second context for the portion of the speech record, wherein the first context has a first probability of correct context and the second context has a second probability of correct context;
processing the portion of the speech record to create a first text translation for the portion of the speech record in the first context, wherein the first text translation has a first probability of correct translation within the first context;
processing the same portion of the speech record to create a second text translation for the portion of the speech record in the second context, wherein the second text translation has a second probability of correct translation within the second context;
processing the first probability of correct translation and the first probability of correct context, to produce a first probability;
processing the second probability of correct translation within the second context and the second probability of correct context, to produce a second probability;
selecting the first translation as the correct translation when the first probability is greater than the second probability; and
selecting the second translation as the correct translation when the second probability is greater than the first probability.

9. The non-transitory computer readable medium of claim 8 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is an identity of a speaker.

10. The non-transitory computer readable medium of claim 8 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is a location of a speaker.

11. The non-transitory computer readable medium of claim 8 , wherein the portion of the speech record is a portion of a sentence and the context for the portion of the speech record is a position of the portion of the speech record within the sentence.

12. The non-transitory computer readable medium of claim 8 , wherein the portion of the speech record is a portion of a call and the context for the portion of the speech record is a position of the portion of the speech record within the call.

13. The non-transitory computer readable medium of claim 8 , wherein the portion of the speech record is a portion of a conversation and the context for the portion of the speech record is a position of the portion of the speech record within the conversation.

14. The non-transitory computer readable medium of claim 8 , wherein the first probability of correct translation within the first context is multiplied by the first probability of correct context, resulting in the first probability.

15. A processing system comprising:
processing circuitry; and
a memory device in communication with the processing circuitry, the memory device having computer-executable instructions stored thereon that, when executed by the processing circuitry, instruct the processing circuitry to:
receive a portion of a speech record from an audio source, the speech record having been produced by the audio source and transferred to the processing circuitry;
determine, from metadata accompanying the portion of the speech record, a first context and a second context for the portion of the speech record, wherein the first context has a first probability of correct context and the second context has a second probability of correct context;
process the portion of the speech record to create a first text translation for the portion of the speech record in the first context, wherein the first text translation has a first probability of correct translation within the first context;
process the same portion of the speech record to create a second text translation for the portion of the speech record in the second context, wherein the second text translation has a second probability of correct translation within the second context;
process the first probability of correct translation and the first probability of correct context, to produce a first probability;
process the second probability of correct translation within the second context and th

Assignees

• Verint Systems Ltd

Inventors

• Brand Michael

Classifications

• G10L15/26 Primary

  Speech to text systems (G10L15/08 takes precedence) · CPC title

• G10L15/1815 Primary

  Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning · CPC title