Generalized phrases in automatic speech recognition systems

• US 9,817,813 B2
• Filed: 01/08/2014
• Issued: 11/14/2017
• Est. Priority Date: 01/08/2014
• Status: Active Grant

• Current Assignee: Genesys Telecommunications Laboratories Incorporated (Genesys Cloud Services Incorporated)
• Original Assignee: Genesys Telecommunications Laboratories Incorporated (Genesys Cloud Services Incorporated)
• Inventors: Faizakof, Avraham, Lev-Tov, Amir, Ollinger, David, Konig, Yochai
• Primary Examiner(s): Riley, Marcus T
• Application Number:
  US14/150,628
• Publication Number:
  US 20150194149A1
• Time in Patent Office: 1,406 Days
• Field of Search: None
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/193 Formal grammars, e.g. finite state automata, context free grammars or word networks

Abstract

A method for generating a suggested phrase having a similar meaning to a supplied phrase in an analytics system includes: receiving, on a computer system comprising a processor and memory storing instructions, the supplied phrase, the supplied phrase including one or more terms; identifying, on the computer system, a term of the phrase belonging to a semantic group; generating the suggested phrase using the supplied phrase and the semantic group; and returning the suggested phrase.

Claims

1. A method comprising:

receiving, on a computer system comprising a processor and memory storing instructions, a supplied phrase, the supplied phrase comprising one or more terms, the supplied phrase being associated with a category of a plurality of categories, each category being associated with a different topic and a plurality of phrases each having a meaning semantically related to the different topic, the plurality of categories being used by an analytics system to perform classifications;

examining one or more terms included in the supplied phrase;

based on examining the one or more terms, determining that a first term of the supplied phrase corresponds to a semantic group;

identifying a second term that is included in the semantic group;

generating, using the supplied phrase and the second term, a suggested phrase having a similar meaning to the supplied phrase, the suggested phrase and the supplied phrase being semantically related to the different topic associated with the category; and

adding the suggested phrase to the category that includes the supplied phrase.

2. The method of claim 1, wherein the semantic group is a formal grammar.

3. The method of claim 2, wherein the formal grammar is configured to match one of an amount of money, a date, a time, a telephone number, a credit card number, a social security number, or a zip code.

4. The method of claim 2, wherein the suggested phrase corresponds to the supplied phrase with the first term replaced with the formal grammar.

5. The method of claim 1, wherein the semantic group comprises a plurality of terms, and wherein the first term is replaced with the second term from the plurality of terms of the semantic group, the second term being different from the first term.

6. The method of claim 5, wherein the semantic group is generated by:

computing differences between each of a plurality of phrases generated by an automatic speech recognition engine, each of the plurality of phrases comprising a plurality of terms;

grouping the plurality of phrases by similarity;

identifying locations of differences between the plurality of phrases; and

defining a generalized semantic group, the generalized semantic group comprising terms at the locations of the differences between the plurality of phrases.

7. The method of claim 5, wherein the second term corresponds to the semantic group comprising the plurality of terms.

8. The method of claim 1, wherein the suggested phrase is supplied as training data to a speech recognition system.

9. The method of claim 1, wherein the analytics system is a speech analytics system.

19. The method of claim 1, wherein the analytics system is configured to performs classifications based on the plurality of categories by:

detecting a phrase of an interaction;

identifying a matching category of the plurality of categories, the matching category being associated with the phrase of the interaction; and

categorizing the interaction as belonging to the matching category.

20. The method of claim 1, wherein adding the suggested phrase to the category that includes the supplied phrase comprises:

displaying the suggested phrase to a user as an alternate option for the supplied phrase;

receiving a user input accepting the suggested phrase as an alternate option; and

in response to receiving the user input, adding the suggested phrase to the category that includes the supplied phrase.

21. The method of claim 6, wherein grouping the plurality of phrases by similarity comprises:

detecting each of the phrases a threshold number of times in a first context; and

in response to the determination, grouping the plurality of phrases associated with a similar context.

10. A system comprising:

a processor; and

a memory, wherein the memory stores instructions that, when executed by the processor, causes the processor to;

receive a supplied phrase, the supplied phrase comprising one or more terms, the supplied phrase being associated with a category of a plurality of categories, each category being associated with a different topic and a plurality of phrases each having a meaning semantically related to the different topic, the plurality of categories being used by an analytics system to perform classifications;

examine one or more terms included in the supplied phrase;

based on examining the one or more terms, determine that a first term of the supplied phrase corresponds to a semantic group;

identify a second term that is included in the semantic group;

generate, using the supplied phrase and the second term, a suggested phrase having a similar meaning to the supplied phrase, the suggested phrase and the supplied phrase being semantically related to the different topic associated with the category; and

configure the analytics system by adding the suggested phrase to the category that includes the supplied phrase.

11. The system of claim 10, wherein the semantic group is a formal grammar.

12. The system of claim 11, wherein the formal grammar is configured to match one of an amount of money, a date, a time, a telephone number, a credit card number, a social security number, or a zip code.

13. The system of claim 11, wherein the memory stores instructions to generate the suggested phrase by replacing the first term in the supplied phrase with the formal grammar.

14. The system of claim 10, wherein the semantic group comprises a plurality of terms, and wherein the memory stores instructions to replace the first term with the second term from the plurality of terms of the semantic group, the second term being different from the first term.

15. The system of claim 14, wherein the memory stores instructions to cause the processor to generate the semantic group generated by:

computing differences between each of a plurality of phrases generated by an automatic speech recognition engine, each of the plurality of phrases comprising a plurality of terms;

grouping the plurality of phrases by similarity;

identifying locations of differences between the plurality of phrases; and

defining a generalized semantic group, the generalized semantic group comprising terms at the locations of the differences between the plurality of phrases.

16. The system of claim 14, wherein the second term corresponds to the semantic group comprising the plurality of terms.

17. The system of claim 10, wherein the suggested phrase is supplied as training data to a speech recognition system.

18. The system of claim 10, wherein the processor and memory are components of a speech analytics system.