Semantic interpretation using user gaze order

• US 8,571,851 B1
• Filed: 12/31/2012
• Issued: 10/29/2013
• Est. Priority Date: 12/31/2012
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Tickner, Simon, Cohen, Richard Z.
• Primary Examiner(s): SPOONER, LAMONT M
• Application Number:
  US13/731,508
• Time in Patent Office: 302 Days
• Field of Search: 704/1, 704/9, 704/10, 704/235, 704/257, 704/275
• US Class Current: 704/9
• CPC Class Codes: G06F 40/30 Semantic analysis
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/24 Speech recognition using non-acoustical features

Abstract

Methods, and systems, including computer programs encoded on computer-readable storage mediums, including a method for performing semantic interpretation using gaze order. The method includes obtaining data identifying a sequence of gaze attention dwell positions, obtaining a semantic description of elements displayed on a visual display, obtaining a transcription of an utterance, correlating the gaze attention dwell positions with the semantic description of elements to generate a sequence of one or more of the elements, performing semantic interpretation of at least one term included in the transcription based at least on the sequence of the elements, and outputting a result of performing the semantic interpretation of the at least one term.

Claims

1. A computer-implemented method for semantically interpreting speech recognition results using gaze tracking data, the method comprising:

obtaining data identifying a sequence of gaze attention dwell positions and a gaze order associated with each gaze attention dwell position, wherein a gaze attention dwell position corresponds to a position on a user interface in which a duration of attention on a single element satisfies a minimum dwell time threshold;

after obtaining the data identifying the sequence of gaze attention dwell positions and the gaze order associated with each gaze attention dwell position, receiving audio data encoding an utterance;

obtaining a transcription of the utterance;

correlating the gaze attention dwell positions and the gaze order associated with each gaze attention dwell position with a semantic description of elements displayed on a visual display to identify one or more particular elements and a respective gaze order associated with each particular element;

performing semantic interpretation of at least one term included in the transcription based at least on the gaze order associated with the particular elements; and

outputting a result of performing the semantic interpretation of the at least one term.

2. The method of claim 1, further comprising:

receiving data identifying a respective dwell time for each of the gaze attention dwell positions; and

weighting each of the particular elements according to the dwell time identified for the respective element.

3. The method of claim 1, further comprising:

obtaining a map of gaze attention positions over time, wherein obtaining data identifying a sequence of gaze attention dwell positions comprises comparing determining whether a respective dwell time associated with each gaze attention dwell position satisfies a minimum dwell time threshold.

4. The method of claim 1, wherein performing semantic interpretation comprises determining that a particular term included in the transcription is associated with one of the elements in the one or more particular elements.

5. The method of claim 1, wherein performing semantic interpretation comprises:

determining that a particular element displayed on the visual display is not one of the elements in the sequence of elements; and

eliminating the particular element as a candidate interpretation of the particular term based at least on determining that the particular element is not one of the elements in the one or more particular elements.

6. The method of claim 1, wherein performing semantic interpretation comprises:

determining that a first term included in the transcription is a substitute for a first element in the one or more particular elements;

determining that a second term that occurs after the first term in the transcription is associated with a second element that occurs later in order than the first element;

selecting the first element or the second element as candidate interpretations of the first term and the second term, respectively, based on determining that (i) the first term included in the transcription is a substitute for the first element and (ii) the second term included in the transcription is a substitute for the second element.

7. The method of claim 1, wherein performing semantic interpretation comprises:

determining that a first term included in the transcription is a substitute for a first element of the one or more particular elements; and

eliminating a second element that occurs earlier than the first element in the sequence of elements as a candidate interpretation of a second term that occurs after the first term in the transcription based at least on determining that the first term included in the transcription is a substitute for the first element in the sequence of elements.

8. The method of claim 1, wherein performing semantic interpretation comprises determining that a particular term included in the transcription is a substitute for one of the elements of the one or more particular elements.

9. The method of claim 1, wherein performing semantic interpretation comprises determining that one of the elements of the one or more particular elements is a concept or entity that is associated with a particular term included in the transcription.

10. The method of claim 1, wherein performing semantic interpretation comprises determining that a particular element of the one or more particular elements that has a largest dwell time is associated with a particular term included in the transcription.

11. The method of claim 1, wherein performing semantic interpretation comprises determining that a last element of the one or more particular elements is associated with a particular term included in the transcription.

12. The method of claim 1, wherein the semantic interpretation is performed further based on a user language setting.

13. The method of claim 1, wherein outputting a result of performing the semantic interpretation comprises outputting the transcription of the utterance, wherein the transcription or the term in the transcription is tagged with at least one element of the one or more particular elements.

14. The method of claim 1, wherein correlating the gaze attention dwell positions with the semantic description of elements comprises matching one or more of the gaze attention dwell positions with one or more positions of elements indicated in the semantic description.

15. The method of claim 1, wherein correlating the gaze attention dwell positions with the semantic description of elements comprises overlaying the gaze attention dwell positions with the semantic description of the elements.

16. A system comprising:

one or more computers; and

a computer readable storage medium storing computer software instructions executable by the one or more computers to perform operations comprising;

obtaining data identifying a sequence of gaze attention dwell positions and a gaze order associated with each gaze attention dwell position;

obtaining a transcription of an utterance;

correlating the gaze attention dwell positions and the gaze order associated with each gaze attention dwell position with a semantic description of elements displayed on a visual display to identify one or more particular elements and a respective gaze order associated with each particular element;

performing semantic interpretation of at least one term included in the transcription based at least on the gaze order associated with particular elements; and

outputting a result of performing the semantic interpretation of the at least one term.

17. The system of claim 16, wherein the operations comprise:

receiving data identifying a respective dwell time for each of the gaze attention dwell positions; and

weighting each of the one or more particular elements according to the dwell time identified for the respective element.

18. A computer-readable storage device storing software comprising instructions executable by one or more computers which, upon such execution, cause the one or more computers to perform operations comprising:

obtaining data identifying a sequence of gaze attention dwell positions and a gaze order associated with each gaze attention dwell position;

obtaining a transcription of an utterance;

correlating the gaze attention dwell positions and the gaze order associated with each gaze attention dwell position with a semantic description of elements displayed on a visual display to identify one or more particular elements and a respective gaze order associated with each particular element;

performing semantic interpretation of at least one term included in the transcription based at least on the gaze order associated with particular elements; and

outputting a result of performing the semantic interpretation of the at least one term.

19. The device of claim 18, wherein a gaze attention dwell position corresponds to a position on a user interface in which a duration of attention on a single element satisfies a minimum dwell time threshold.

20. The device of claim 18, wherein the semantic description comprises a description of elements displayed on the virtual display, and one or more positions on the virtual display associated with each element.