Script compliance in spoken documents

Patent metadata

Field	Value
Publication number	US-2016133256-A1
Application number	US-201414538833-A
Country	US
Kind code	A1
Filing date	Nov 12, 2014
Priority date	Nov 12, 2014
Publication date	May 12, 2016
Grant date	—

Abstract

A method, computerized apparatus and computer program product for determining script compliance in interactions, the method comprising: receiving one or more indexed audio interaction; receiving a text representing a script; automatically extracting two or more key terms from the script; automatically generating a query representing the script, comprising: receiving one or more constraint associated with the at least two key terms; and determining spotted key terms of the key terms that appear in the indexed audio interactions; determining complied constraints; and determining a relevance score for each of the indexed audio interactions, based on the spotted key terms and the complied constraints.

First claim

What is claimed is:

1 . A computerized method for determining compliance of audio interactions with a script, comprising:
receiving at least one indexed audio interaction; receiving a text representing a script; automatically extracting at least two key terms from the script; automatically generating a query representing the script, comprising:
receiving at least one constraint associated with the at least two key terms; and
determining spotted key terms of the at least two key terms that appear in the at least one indexed audio interaction;
determining complied constraints of the at least one constraint; and
determining a relevance score for the at least one indexed audio interaction, based on the spotted key terms and the complied constraints.

2 . The method of claim 1 , wherein the at least one constraint refers to order and distance between the at least two key terms in the indexed audio interaction, wherein the distance relates to time difference or number of words difference between the at least two key terms.

3 . The method of claim 1 , wherein the at least one indexed audio interaction comprises a multiplicity of indexed audio interactions, and further comprising filtering indexed audio interactions based on the relevance score.

4 . The method of claim 3 further comprising:
sorting the multiplicity of indexed audio interactions based on the relevance score, and
filtering from the multiplicity of indexed audio interactions those interactions having the highest or lowest relevance scores.

5 . The method of claim 1 , wherein the query comprises a weight associated with at least one of the at least two key terms.

6 . The method of claim 5 , wherein determining the relevant score comprises:
determining at least one group of key terms, wherein a key term in the at least one group appears in the at least one interaction, and the key term is connected to another key term in the at least one group if the key term and the other key term appear in one constraint complied with by the interaction; determining a score for the at least one group as a function of weights associated with key terms that belong to the at least one group; and determining the relevance score as a maximal relevance score associated with any group of the at least one group.

7 . The method of claim 1 , wherein extracting the at least two key terms comprises:
discarding irrelevant terms; determining a recognizability score for a word in the script; determining a uniqueness score for the word in the script; combining the recognizability score and the uniqueness score into a word score; and determining that the word is a key term, subject to the word score exceeding a threshold.

8 . The method of claim 7 , wherein the uniqueness score is determined using a Term Frequency-Inverse Document Frequency (TF-IDF) algorithm.

9 . The method of claim 1 , wherein extracting the at least two key terms comprises:
discarding irrelevant terms; determining a recognizability score for a word in the script; determining a uniqueness score for the word in the script; and determining that the word is a key term, subject to the recognizability score exceeding a first threshold, or the uniqueness score exceeding a second threshold.

10 . The method of claim 1 , further comprising:
receiving an audio interaction; decoding speech within the audio interaction; and indexing the decoded speech to obtain the at least one indexed audio interaction.

11 . A computerized apparatus having a processor, the processor being adapted to perform the steps of:
receiving at least one indexed audio interaction; receiving a text representing a script; automatically extracting at least two key terms from the script; automatically generating a query representing the script, comprising:
receiving at least one constraint associated with the at least two key terms; and
determining spotted key terms of the at least two key terms that appear in the at least one indexed audio interaction;
determining complied constraints of the at least one constraint; and
determining a relevance score for the at least one indexed audio interaction, based on the spotted key terms and the complied constraints.

12 . The computerized apparatus of claim 11 , wherein the at least one constraint refers to order and distance between the at least two key terms in the indexed audio interaction, wherein the distance relates to time difference or number of words difference between the at least two key terms.

13 . The computerized apparatus of claim 11 , wherein the at least one indexed audio interaction comprises a multiplicity of indexed audio interactions, and further comprising filtering indexed audio interactions based on the relevance score.

14 . The computerized apparatus of claim 13 , wherein the processor is further adapted to perform the steps of:
sorting the multiplicity of indexed audio interactions based on the relevance score, and filtering from the multiplicity of indexed audio interactions those interactions having the highest or lowest relevance scores.

15 . The computerized apparatus of claim 11 , the relevance score is determined by:
determining at least one group of key terms, wherein a key term in the at least one group appears in the at least one interaction, and the key term is connected to another key term in the at least one group if the key term and the other key term appear in one constraint complied with by the interaction; determining a score for the at least one group as a function of weights associated with key terms that belong to the at least one group; and determining the relevance score as a maximal relevance score associated with any group of the at least one group.

16 . The computerized apparatus of claim 11 , wherein extracting the at least two key terms comprises:
discarding irrelevant terms; determining a recognizability score for a word in the script; determining a uniqueness score for the word in the script; combining the recognizability score and the uniqueness score into a word score; and determining that the word is a key term, subject to the word score exceeding a threshold.

17 . The computerized apparatus of claim 11 , wherein extracting the at least two key terms comprises:
discarding irrelevant terms; determining a recognizability score for a word in the script; determining a uniqueness score for the word in the script; and determining that the word is a key term, subject to the recognizability score exceeding a first threshold, or the uniqueness score exceeding a second threshold.

18 . The computerized apparatus of claim 11 , wherein the processor is further adapted to perform the steps of:
receiving the audio interaction; decoding speech within the audio interaction; and indexing the decoded speech to obtain the at least one indexed audio interaction.

19 . The computerized apparatus of claim 11 , wherein the processor is further adapted to perform the steps of:
receiving a multiplicity of audio interactions; decoding speech within each of the multiplicity of audio interactions; indexing the decoded speech to obtain indexed audio interactions; repeating the steps of claim 11 for each indexed audio interaction; and filtering indexed audio interactions for which the score is above or below a predetermined value.

20 . A computer program product comprising a computer readable storage medium retaining program instructions, which program instructions wh

Assignees

• Nice Systems Ltd

Inventors

• Lembersky Gennadi
• Nissan Maor
• Bretter Ronny

Classifications

• H04M3/5175

  Call or contact centers supervision arrangements · CPC title

• H04M3/42221

  Conversation recording systems (at the subscriber's set H04M1/656) · CPC title

• G10L25/48

  specially adapted for particular use · CPC title

• G10L15/197

  Probabilistic grammars, e.g. word n-grams · CPC title

• G10L15/1815

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