Methods for statistical analysis of speech

• US 20080208582A1
• Filed: 01/29/2008
• Published: 08/28/2008
• Est. Priority Date: 09/27/2002
• Status: Active Grant

• Current Assignee: CallMiner, Inc.
• Original Assignee: CallMiner, Inc.
• Inventors: Gallino, Jeffrey A.
• Granted Patent: US 8,583,434 B2
• Time in Patent Office: Days
• US Class Current: 704/251
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/14 using statistical models, e.g. Hidden Markov Models [HMMs] G10L15/18 takes precedence
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/0636 Threshold criteria for the updating

Abstract

Computer-implemented methods and apparatus are provided to facilitate the recognition of the content of a body of speech data. In one embodiment, a method for analyzing verbal communication is provided, comprising acts of producing an electronic recording of a plurality of spoken words; processing the electronic recording to identify a plurality of word alternatives for each of the spoken words, each of the plurality of word alternatives being identified by comparing a portion of the electronic recording with a lexicon, and each of the plurality of word alternatives being assigned a probability of correctly identifying a spoken word; loading the word alternatives and the probabilities to a database for subsequent analysis; and examining the word alternatives and the probabilities to determine at least one characteristic of the plurality of spoken words.

Claims

1. A method for analyzing verbal communication, the method comprising acts of:

(A) producing an electronic recording of a plurality of communications, each communication comprising a plurality of words spoken by one or more individuals;

(B) processing the electronic recording to identify a plurality of word alternatives for each of the spoken words, each of the plurality of word alternatives being identified by comparing a portion of the electronic recording to a lexicon, each of the plurality of word alternatives being assigned a probability of correctly identifying a spoken word;

(C) loading the word alternatives and the probabilities to an electronic file storage for analysis; and

(D) examining the word alternatives and the probabilities to determine at least one characteristic of the plurality of spoken words other than a word alternative having a highest probability of correctly identifying a spoken word.

2. The method of claim 1, wherein the act (D) comprises determining a frequency at which a first word alternative occurs within the plurality of communications.

3. The method of claim 2, wherein the act (D) further comprises determining a frequency at which a second word alternative occurs within the plurality of communications, and comparing the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs.

4. The method of claim 3, wherein the act (D) comprises comparing the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs in a plurality of time periods.

5. The method of claim 1, wherein the act (D) further comprises:

(D1) receiving input from a user via a graphical user interface (GUI) defining a pattern, the pattern comprising a representation of one or more concepts;

(D2) comparing the pattern to the word alternatives; and

(D3) causing an indication of a match between the pattern and any one or more of the word alternatives to be stored in an electronic file storage.

6. The method of claim 5, wherein the word alternatives comprise a plurality of data segments, (D1) further comprises receiving input from the user defining a threshold score and a plurality of patterns each assigned a pattern score, (D2) further comprises comparing the plurality of patterns to the data segments, and (D3) further comprises assigning a pattern score to a data segment when one of the plurality of patterns matches the data segment, and causing an indication of a segment match to be stored when a sum of the pattern scores assigned to the data segment meets or exceeds the threshold score.

7. The method of claim 5, wherein the word alternatives comprise a plurality of data segments, each data segment comprising one of the plurality of communications.

8. The method of claim 7, further comprising an act (E) of analyzing the data segments for which an indication of a segment match was caused to be stored in (D3).

9. The method of claim 8, wherein each data segment comprises a communication, and the act (E) comprises listening to a recording of one or more of the communications.

10. At least one computer-readable medium having instructions recorded thereon which, when executed by a computer implemented in a system comprising a speech recognition tool and an electronic file storage, the speech recognition tool being operable to process an electronic recording of at least one word to recognize a plurality of alternatives for the at least one word, the electronic file storage storing at least a portion of the plurality of alternatives, perform a method of analyzing verbal communication, the method comprising acts of:

(A) providing a graphical user interface (GUI) through which a user may provide input defining a pattern, the pattern comprising a representation of one or more concepts;

(B) comparing the pattern to the alternatives stored in the electronic file storage to determine a match between the pattern and any one or more of the alternatives; and

(C) causing an indication of a match between the pattern and any one or more of the alternatives to be stored in an electronic file storage.

11. The at least one computer-readable medium of claim 10, wherein the electronic file storage stores a plurality of data segments, each data segment comprising a plurality of word alternatives, and wherein the act (A) further comprises providing a GUI through which a user may provide input defining a threshold score and a plurality of patterns each assigned a pattern score, the act (B) further comprises comparing the plurality of patterns to the data segments and assigning a pattern score to a data segment when one of the plurality of patterns matches that data segment, and the act (C) further comprises causing an indication of a segment match to be stored when the sum of all pattern scores assigned to the segment meets or exceeds the threshold score.

12. At least one computer-readable medium having instructions encoded thereon which, when executed by one or more computers, cause the one or more computers to perform a method for analyzing verbal communication, the method comprising acts of:

(A) producing an electronic recording of a plurality of communications, each communication comprising a plurality of words spoken by one or more individuals;

13. The at least one computer-readable medium of claim 12, wherein the act (D) comprises determining a frequency at which a first word alternative occurs within the plurality of communications.

14. The at least one computer-readable medium of claim 13, wherein the act (D) further comprises determining a frequency at which a second word alternative occurs within the plurality of communications, and comparing the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs.

15. The at least one computer-readable medium of claim 14, wherein the act (D) comprises comparing the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs in a plurality of time periods.

16. The at least one computer-readable medium of claim 12, wherein the act (D) further comprises:

(D1) receiving input from a user via a graphical user interface (GUI) defining a pattern, the pattern comprising a representation of one or more concepts;

(D2) comparing the pattern to the word alternatives; and

(D3) causing an indication of a match between the pattern and any one or more of the word alternatives to be stored in an electronic file storage.

17. The at least one computer-readable medium of claim 16, wherein the word alternatives comprise a plurality of data segments, (D1) further comprises receiving input from the user defining a threshold score and a plurality of patterns each assigned a pattern score, (D2) further comprises comparing the plurality of patterns to the data segments, and (D3) further comprises assigning a pattern score to a data segment when one of the plurality of patterns matches the data segment, and causing an indication of a segment match to be stored when a sum of the pattern scores assigned to the data segment meets or exceeds the threshold score.

18. The at least one computer-readable medium of claim 16, wherein the word alternatives comprise a plurality of data segments, each data segment comprising one of the plurality of communications.

19. The at least one computer-readable medium of claim 18, further comprising an act (E) of analyzing the data segments for which an indication of a segment match was caused to be stored in (D3).

20. The at least one computer-readable medium of claim 19, wherein each data segment comprises a communication, and wherein the act (E) comprises listening to one or more of the communications.

21. An apparatus operable to analyze verbal communication, comprising:

a recording controller operable to produce an electronic recording of a plurality of communications, each communication comprising a plurality of words spoken by one or more individuals;

an identification controller operable to process the electronic recording to identify a plurality of word alternatives for each of the spoken words, each of the plurality of word alternatives being identified by comparing a portion of the electronic recording to a lexicon, each of the plurality of word alternatives being assigned a probability of correctly identifying a spoken word;

an electronic file storage operable to store the word alternatives and the probabilities; and

an analysis controller operable to examine the word alternatives and the probabilities stored in the electronic file storage to determine at least one characteristic of the plurality of spoken words other than a word alternative having a highest probability of correctly identifying a spoken word.

22. The apparatus of claim 21, wherein the analysis controller is further operable to determine a frequency at which a first word alternative occurs within the plurality of communications.

23. The apparatus of claim 22, wherein the analysis controller is further operable to determine a frequency at which a second word alternative occurs within the plurality of communications, and compare the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs.

24. The apparatus of claim 23, wherein the analysis controller is further operable to compare the frequency at which the first word alternative occurs with the frequency at which the second word alternative occurs in a plurality of time periods.

25. The apparatus of claim 21, wherein the analysis controller is further operable to:

receive input from a user via a graphical user interface (GUI) defining a pattern, the pattern comprising a representation of one or more concepts;

compare the pattern to the word alternatives; and

cause an indication of a match between the pattern and any one or more of the word alternatives to be stored in an electronic file storage.

26. The apparatus of claim 25, wherein the word alternatives comprise a plurality of data segments, and wherein the analysis controller is further operable to receive input from the user defining a threshold score and a plurality of patterns each assigned a pattern score, compare the plurality of patterns to the data segments, assign a pattern score to a data segment when one of the plurality of patterns matches the data segment, and cause an indication of a segment match to be stored when a sum of the pattern scores assigned to the data segment meets or exceeds the threshold score.

27. The apparatus of claim 25, wherein the word alternatives comprise a plurality of data segments, and wherein the analysis controller is further operable to analyze the data segments for which an indication of a segment match is caused to be stored.

28. The apparatus of claim 27, wherein each data segment comprises a communication, and further comprising a playback controller operable to enable a user to listen to a recording of one or more of the communications.