Data driven speech enabled self-help systems and methods of operating thereof

Patent metadata

Field	Value
Publication number	US-10515150-B2
Application number	US-201514799369-A
Country	US
Kind code	B2
Filing date	Jul 14, 2015
Priority date	Jul 14, 2015
Publication date	Dec 24, 2019
Grant date	Dec 24, 2019

Abstract

A method for configuring an automated, speech driven self-help system based on prior interactions between a plurality of customers and a plurality of agents includes: recognizing, by a processor, speech in the prior interactions between customers and agents to generate recognized text; detecting, by the processor, a plurality of phrases in the recognized text; clustering, by the processor, the plurality of phrases into a plurality of clusters; generating, by the processor, a plurality of grammars describing corresponding ones of the clusters; outputting, by the processor, the plurality of grammars; and invoking configuration of the automated self-help system based on the plurality of grammars.

First claim

What is claimed is:

1. A method for configuring an automated, speech driven interactive response system based on prior interactions between a plurality of customers and a plurality of agents, the method comprising:
recognizing, by a processor, speech in the prior interactions between customers and agents to generate recognized text;
detecting, by the processor, a plurality of phrases in the recognized text;
clustering, by the processor, the plurality of phrases into a plurality of clusters, each cluster comprising a plurality of phrases corresponding to a topic;
receiving, by the processor, an input phrase to configure the automated interactive response system to perform an action in response to customer speech comprising the input phrase;
determining, by the processor, whether the input phrase belongs to one or more matching clusters of the plurality of clusters;
in response to determining that the input phrase belongs to one or more matching clusters:
identifying, by the processor, a plurality of suggested phrases from the one or more matching clusters, the suggested phrases comprising a generalized phrase corresponding to the input phrase, the generalized phrase comprising a wildcard term replacing one or more words or terms of the input phrase; and
outputting, by the processor, the suggested phrases associated with the one or more matching clusters, the suggested phrases being semantically similar to the input phrase;

in response to determining that the input phrase does not belong to one or more matching clusters:
computing a semantic distance between the input phrase and a plurality of medoids of the plurality of clusters; and
outputting the suggested phrases associated with a cluster of the plurality of clusters wherein the semantic distance between the input phrase and a medoid of the cluster is below a threshold distance;

generating, by the processor, a grammar of a plurality of grammars, the grammar being configured to match the input phrase and at least one of the suggested phrases in accordance with the generalized phrase comprising the wildcard term, the grammar being associated with the action to trigger the automated interactive response system to perform the action in response to detecting the input phrase or the at least one of the suggested phrases;
outputting, by the processor, the grammar of the plurality of grammars; and
invoking configuration of the automated interactive response system based on the plurality of grammars by configuring the automated interactive response system to automatically:
receive audio from a customer;
compare the audio to the plurality of grammars to identify an action associated with a grammar matching the audio;
generate a response in accordance with the action associated with the matching grammar; and
transmit the response to the customer.

2. The method of claim 1 , wherein the identifying the plurality of suggested phrases comprises calculating a semantic distance between the input phrase and each of the phrases in each of the one or more matching clusters,
wherein the suggested phrases include phrases having a semantic distance below a threshold semantic distance and exclude phrases having a semantic distance above the threshold semantic distance.

3. The method of claim 1 , wherein the identifying the plurality of suggested phrases comprises: identifying, by the processor, existing word groups and grammars associated with the input phrase.

4. A system comprising:
a processor; and
memory storing instructions that, when executed on the processor, cause the processor to:
recognize speech in a plurality of prior interactions between customers and agents to generate recognized text;
detect a plurality of phrases in the recognized text;
cluster the plurality of phrases into a plurality of clusters, each cluster comprising a plurality of phrases corresponding to a topic;
receive an input phrase to configure an automated interactive response system to perform an action in response to customer speech comprising the input phrase;
determine whether the input phrase belongs to one or more matching clusters of the plurality of clusters;
in response to determining that the input phrase belongs to one or more matching clusters:
identify a plurality of suggested phrases from the one or more matching clusters, the suggested phrases comprising a generalized phrase corresponding to the input phrase, the generalized phrase comprising a wildcard term replacing one or more words or terms of the input phrase; and
output the suggested phrases associated with the one or more matching clusters, the suggested phrases being semantically similar to the input phrase;

in response to determining that the input phrase does not belong to one or more matching clusters:
compute a semantic distance between the input phrase and a plurality of medoids of the plurality of clusters; and
output the suggested phrases associated with a cluster of the plurality of clusters wherein the semantic distance between the input phrase and a medoid of the cluster is below a threshold distance;

generate a grammar of a plurality of grammars, the grammar being configured to match the input phrase and at least one of the suggested phrases in accordance with the generalized phrase comprising the wildcard term, the grammar being associated with the action to trigger the automated interactive response system to perform the action in response to detecting the input phrase or the at least one of the suggested phrases;
output the grammar of the plurality of grammars; and
invoke configuration of an automated interactive response system based on the plurality of grammars, the configuration of the automated interactive response system comprising configuring the automated interactive response system to automatically:
receive audio from a customer;
compare the audio to the plurality of grammars to identify an action associated with a grammar matching the audio;
generate a response in accordance with the action associated with the matching grammar; and
transmit the response to the customer.

5. The system of claim 4 , wherein the memory further stores instructions that, when executed on the processor, cause the processor to identify the plurality of suggested phrases by calculating a semantic distance between the input phrase and each of the phrases in each of the one or more matching clusters,
wherein the suggested phrases include phrases having a semantic distance below a threshold semantic distance and exclude phrases having a semantic distance above the threshold semantic distance.

6. The system of claim 4 , wherein the memory further stores instructions that, when executed on the processor, cause the processor to identify the plurality of suggested phrases by identifying, by the processor, existing word groups and grammars associated with the input phrase.

7. The method of claim 1 , wherein the automated interactive response system is an interactive voice response (IVR) system.

8. The method of claim 1 , wherein the automated interactive response system is an interactive media response (IMR) system.

9. The system of claim 4 , wherein the automated interactive response system is an interactive voice response (IVR) system.

10. The system of claim 4 , wherein the automated interactive response system is an interactive media response (IMR) system.

Assignees

• Genesys Telecommunications Laboratories Inc

Inventors

• Lev Yoni
• Tapuhi Tamir
• Faizakof Avraham
• Lev-Tov Amir
• Konig Yochai

Classifications

• G10L15/32

  Multiple recognisers used in sequence or in parallel; Score combination systems therefor, e.g. voting systems · CPC title

• G10L15/19 Primary

  Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules · CPC title

• G10L15/1822

  Parsing for meaning understanding · CPC title

• G06F40/289 Primary

  Phrasal analysis, e.g. finite state techniques or chunking · CPC title

• G10L2015/223

  Execution procedure of a spoken command · CPC title