Speech recognition confidence level display

• US 6,006,183 A
• Filed: 12/16/1997
• Issued: 12/21/1999
• Est. Priority Date: 12/16/1997
• Status: Expired due to Fees

• Current Assignee: International Business Machines Corporation
• Original Assignee: International Business Machines Corporation
• Inventors: Vergo, John George, Lai, Jennifer Ceil
• Primary Examiner(s): Hudspeth, David R.
• Assistant Examiner(s): Smits, Talivaldis Ivars
• Application Number:
  US08/991,264
• Time in Patent Office: 735 Days
• Field of Search: 704/235, 704/251
• US Class Current: 704/235
• CPC Class Codes: G10L 15/22 Procedures used during a speech recognition process, e.g. man-machine dialogue
• CPC Class Codes: G10L 2015/221 Announcement of recognition results

Abstract

A speech recognition computer system and method indicates the level of confidence that a speech recognizer has in it recognition of one or more displayed words. The system and method allow for the rapid identification of speech recognition errors. A plurality of confidence levels of individual recognized words may be visually indicated. Additionally, the system and method allow the user of the system to select threshold levels to determine when the visual indication occurs.

Claims

1. A speech recognition system comprising:

a speech recognizer for translating speech into text, said text being one or more sentences, said speech recognizer further comprising a confidence level scorer for assigning one of a selected number of possible scores for each word of said one or more sentences, said score being a confidence measure that each said word has been recognized correctly;

a user interface for displaying said text, said each word having display properties based on said scores and providing a visual indication of said confidence measure for said each word, wherein said display properties include a default display property and two or more other display properties; and

a user control which allows a user of said speech recognition system to select said number and define a score threshold for displaying said default display property.

2. A speech recognition system according to claim 1, wherein said default display property is normal text.

3. A speech recognition system according to claim 1, wherein said each word is displayed with one of said two or more other display properties when said confidence measure is below said score threshold, thereby indicating a possible error.

4. A speech recognition system according to claim 3, wherein said score threshold is an upper limit of said number of thresholds selected by said user, said thresholds defining a plurality of sub-ranges, each of said display properties corresponding to one of said defined sub-ranges.

5. A speech recognition system according to claim 1, wherein said each word is displayed with said default display property when said confidence measure is above said score threshold level.

6. A speech recognition system according to claim 1, wherein said default display property corresponds to scores above said threshold score and each of said two or more other display properties correspond to different ones of said selected number of possible scores.

7. A speech recognition system comprising:

a speech recognizer for translating speech into text, said text being one or more sentences, said speech recognizer comprising a confidence level scorer for assigning one of a user selected number possible scores for each word of said one or more sentences;

a user interface for displaying said text based on said scores, wherein said scores cause said each word to be displayed in accordance with one of a user selected number of indications, each of said indications indicating into which of a selected number of sub-ranges said scores fall; and

a user control which allows a user of said speech recognition system to select said number of sub-ranges and boundaries of each sub-range.

8. A speech recognition system according to claim 7, wherein each of said indications is a different color.

9. A speech recognition system according to claim 7, wherein each of said indications is at least one different font attribute selected from the group consisting of font type, point size, bold, italics, underline, double underline, capitalization, flashing and blinking.

10. A speech recognition system according to claim 7, wherein each of indications is one of a different shade of a single color or a different shade of gray.

11. A speech recognition system according to claim 10, wherein said user control enables said user to select one of a color map or a gray scale map to identify said sub-ranges.

12. A speech recognition system comprising:

a speech recognizer for translating speech into text, said text being one or more sentences, said speech recognizer comprising a confidence level scorer assigning one of a user selected number possible scores for each word of said one or more sentences;

a user interface for displaying said text in accordance with said scores, wherein said scores are mapped to a continuous range of colors, said each word being displayed in said mapped color, thereby providing a different color display for different respective words having different said scores; and

a user control allowing a user to select a score threshold, scores beyond said score threshold being displayed with a default display property.

13. A speech recognition system according to claim 12, further comprising color maps for said range of colors, wherein said user of said speech recognition system selects one of said color maps.

14. A method of speech recognition comprising:

translating input speech into text, said text being one or more sentences;

assigning one of at least three possible confidence level scores for each word of said one or more sentences, said score being a confidence measure that said each word has been recognized correctly;

selecting by a user a number of score sub-ranges and a score limit; and

displaying said text, each said word having one of at least three display properties corresponding to a score falling within one of said score sub-ranges or beyond said score limit, said display properties providing a visual indication of said confidence measure for said each word.

15. A method of speech recognition according to claim 14, wherein said at least three display properties include a default display property corresponding to scores beyond said score limit and two or more other display properties.

16. A method of speech recognition according to claim 15, wherein said default display property is normal text.

17. The method of speech recognition according to claim 15, wherein said one or more words is displayed with one of said two or more other display properties when said confidence measure of said one or more words is below said score limit.

18. The method of speech recognition according to claim 15, further comprising the step of:

displaying said each word with said default display property when said confidence measure is above said score limit.

19. A method of speech recognition comprising:

translating input speech into text, said text being one or more sentences;

assigning one of a user selected number confidence level scores for each word of said one or more sentences; and

displaying said text, wherein said scores cause said each word to be displayed in accordance with one of a user selected number of indications, depending upon which of a plurality of sub-ranges each of said scores fall into, said sub-ranges being selected by said user.

20. The method of speech recognition according to claim 19, further comprising the step of:

providing a different color for each of said plurality of indications.

21. The method of speech recognition according to claim 19, further comprising the step of:

providing at least one different font attribute for each of said plurality of indications, said different font attribute being selected from the group consisting of font type, point size, bold, italics, underline, double underline, capitalization, flashing and blinking.

22. The method of speech recognition according to claim 19, further comprising the step of:

providing one of a different shade of a single color or a different shade of gray for each of said plurality of indications.

23. A method of speech recognition computer system according to claim 22, further comprising the step of:

said user selecting one of a color map or a gray scale map to identify said sub-ranges.

24. A method of speech recognition comprising:

translating input speech into text, said text being one or more sentences;

assigning one of a user selected number of possible confidence level scores for each word of said one or more sentences; and

displaying said text in accordance with said scores, wherein said scores are mapped to a continuous range of colors, said user defining an upper confidence level being mapped to said continuous range, said each word being displayed in said mapped color, thereby providing a different color display for different respective words having different said scores.

25. The method of speech recognition according to claim 24, further comprising the step of:

selecting a color map for said range of colors.

26. A speech recognition system comprising:

a speech recognizer for translating speech into text, said text being one or more sentences, said speech recognizer comprising a confidence level scorer assigning a confidence level to each word of said one or more sentences;

a confidence level indicator assigning a confidence level indication color to said each word, said color being selected from a range of colors based on said assigned confidence level;

a user interface displaying said text, said each word being displayed with its assigned color, the confidence level of said each word being visually discernable; and

a user control allowing a user to select said range of colors and confidence range limit.

27. A speech recognition system as in claim 26, wherein when said confidence level is above said confidence range limit, the assigned color is a default color.

28. A speech recognition system as in claim 26, wherein when said confidence level is below said confidence range limit, the assigned color is a color selected to indicate an error.

29. A method of speech recognition comprising:

translating input speech into text, said text being one or more sentences;

assigning a confidence level for each word of said text, a confidence level limit being selected by a user; and

mapping said confidence level to a range of display attributes, said range including at least three display attributes the confidence level value subranges being mapped thereby and being user selectable; and

displaying said text, wherein said each word is displayed with a corresponding mapped attribute, such that each word'"'"'s confidence level is visually discernable.

30. A method as in claim 29 wherein said range of display attributes is a range of colors.

31. A method as in claim 29 wherein said range of display attributes is a range of font attributes.