System and Method of Text Zoning

• US 20150066506A1
• Filed: 08/25/2014
• Published: 03/05/2015
• Est. Priority Date: 08/30/2013
• Status: Abandoned Application

• Current Assignee: Verint Systems Limited (Verint Systems Incorporated)
• Original Assignee: Verint Systems Limited (Verint Systems Incorporated)
• Inventors: Romano, Roni, Horesh, Yair, Dreyfuss, Jeremie
• Application Number:
  US14/467,783
• Publication Number:
  US 20150066506A1
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/04 Segmentation; Word boundary detection
• CPC Class Codes: G10L 15/18 using natural language modelling
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A method of zoning a transcription of audio data includes separating the transcription of audio data into a plurality of utterances. A that each word in an utterances is a meaning unit boundary is calculated. The utterance is split into two new utterances at a work with a maximum calculated probability. At least one of the two new utterances that is shorter than a maximum utterance threshold is identified as a meaning unit.

Claims

1. A method of zoning a transcription of audio data, the method comprising:

separating the transcription of audio data into a plurality of utterances;

identifying utterances of the plurality of utterances that are shorter than a predetermined minimum threshold as meaning units;

calculating a probability that each word in an utterance of the plurality of utterances which is longer than the predetermined minimum threshold is a meaning unit boundary;

splitting the utterance longer than the predetermined minimum threshold into two new utterances at a word with a maximum calculated probability; and

identifying at least one of the two utterances that is shorter than a maximum utterance threshold as a meaning unit.

2. The method of claim 1, wherein calculating the probability that each word in the utterance longer than the predetermined minimum threshold is a meaning unit boundary is further based upon at least a linguistic model.

3. The method of claim 2, wherein the linguistic model comprises statistics, distributions, or frequencies of word pairs or word triplets.

4. The method of claim 2, wherein the linguistic model comprises probability of words to form the beginning or end of a meaning unit.

5. The method of claim 2 wherein calculating the probability that each word in the utterance longer than the predetermined minimum threshold is a meaning unit boundary is further based upon an acoustic model.

6. The method of claim 2, further comprising receiving audio data and decoding the audio data to create the transcription of audio data.

7. The method of claim 5, wherein at least the linguistic model is used when decoding the audio data to create the transcription of audio data.

8. The method of claim 1, further comprising applying speech analytics to the meaning unit to identify at least one of context or content of the meaning unit.

9. The method of claim 1, further comprising applying speech analytics to identified meaning units to group the meaning units into call segments.

10. The method of claim 9, further comprising applying speech analytics to the identified meaning units to identify dialog acts within the identified meaning units.

11. The method of claim 1, wherein the predetermined minimum threshold is thirty words.

12. The method of claim 1, further comprising:

selecting utterances of the plurality that are longer than the predetermined minimum threshold for subdivision; and

splitting the selected utterances of the plurality into widows, each window being twice the maximum utterance threshold.

13. The method of claim 12, wherein calculating the probability that each word in an utterance longer than the predetermined minimum threshold is a meaning unit boundary is calculated for each word in each window.

14. The method of claim 13, further comprising applying at least one of a linguistic exception and an acoustic exception to the two new utterances.

15. The method of claim 14, wherein the at least one linguistic exception comprises a minimum meaning unit boundary probability or a minimum meaning unit boundary probability differential.

16. The method of claim 14, wherein the at least one acoustic exception comprises an identification of a pause between adjacent utterances in the transcription of the audio data.

17. A method of zoning, a transcription of audio data, the method comprising:

separating the transcription of audio data into a plurality of utterances

identifying utterances of the plurality of utterances that are shorter than a predetermined minimum threshold as meaning units;

selecting utterances of the plurality of utterances that are longer than the predetermined minimum threshold for subdivision;

splitting the selected utterances into widows, each window being twice a maximum utterance threshold;

calculating a probability that each word in the plurality of windows is a meaning unit boundary based upon at least a linguistic model applied to each of the plurality of windows;

splitting the selected utterances which are longer than the predetermined minimum threshold into two new utterances at a word with a maximum calculated probability; and

identifying at least one of the two new utterances that is shorter than a maximum utterance threshold as a meaning unit.

18. The method of claim 17, wherein the linguistic model comprises probability of words to form the beginning or end of a meaning unit.

19. The method of claim 18, further comprising receiving audio data and decoding the audio data with at least the linguistic model to create the transcription of audio data.

20. The method of claim 19, further comprising applying at least one of a linguistic exception and an acoustic exception to the two new utterances.