Enhancing speech recognition with domain-specific knowledge to detect topic-related content

• US 9,336,776 B2
• Filed: 05/01/2013
• Issued: 05/10/2016
• Est. Priority Date: 05/01/2013
• Status: Active Grant

• Current Assignee: SAP SE
• Original Assignee: SAP SE
• Inventors: Moser, Gerd, Dahlmeier, Daniel, Suleiman, Basem, Roy, Marcus, Schrank, Dominik
• Primary Examiner(s): Godbold, Douglas
• Application Number:
  US13/874,854
• Publication Number:
  US 20140330558A1
• Time in Patent Office: 1,105 Days
• Field of Search: 704/275
• US Class Current: 1/1
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models

Abstract

Methods, systems, and computer-readable storage media for providing action items from audio within an enterprise context. In some implementations, actions include determining a context of audio that is to be processed, providing training data to a speech recognition component, the training data being provided based on the context, receiving text from the speech recognition component, processing the text to identify one or more action items by identifying one or more concepts within the text and matching the one or more concepts to respective transitions in an automaton, and providing the one or more action items for display to one or more users.

Claims

1. A computer-implemented method for providing action items from an audio file within an enterprise context, the method being executed using one or more processors and comprising:

determining, by the one or more processors, a context of the audio file that is to be processed based on a user input indicating a training data selection;

providing, by the one or more processors, training data to a speech recognition component, the training data being in a format recognizable by the speech recognition component and being provided based on the context;

receiving, by the one or more processors, a textual transcript corresponding to the audio file from the speech recognition component;

processing, by the one or more processors, the textual transcript to identify one or more action items by identifying one or more concepts within the textual transcript and matching the one or more concepts to respective transitions in an automaton; and

providing the one or more action items for display to one or more users.

2. The method of claim 1, wherein the automaton comprises a plurality states and one or more transitions, a transition representing a transition between states.

3. The method of claim 1, wherein processing the textual transcript further comprises, for action items of the one or more action items, determining a respective quality score.

4. The method of claim 3, wherein the quality score is determined based on a precision score and a relevance score.

5. The method of claim 4, wherein the precision score is determined based on an accumulated probability of matched transitions of the automaton and a sum of all probabilities of transitions along accepting paths of the automaton.

6. The method of claim 4, wherein the relevance score is determined based on a degree of matching of a path of the action item with respective paths of one or more previously selected action items.

7. The method of claim 3, wherein the one or more actions items are displayed based on respective quality scores.

8. The method of claim 1, wherein the training data comprises domain-specific information provided from a knowledge base.

9. The method of claim 8, wherein the domain-specific information comprises topic-related information and domain-specific terminology.

10. The method of claim 1, wherein the context is determined based on user input.

11. The method of claim 10, wherein the user input comprises user speech provided in the audio.

12. The method of claim 1, further comprising:

receiving user input, the user input indicating selection of an action item of the one or more action items; and

providing the action item to a management component.

13. The method of claim 12, wherein the management component monitors execution of the action item.

14. The method of claim 1, wherein the audio file is provided as real-time audio.

15. The method of claim 1, wherein the audio file is provided as recorded audio.

16. A non-transitory computer-readable storage medium coupled to one or more processors and having instructions stored thereon which, when executed by the one or more processors, cause the one or more processors to perform operations for providing action items from an audio file within an enterprise context, the operations comprising:

determining, by the one or more processors, a context of the audio file that is to be processed based on a user input indicating a training data selection;

providing, by the one or more processors, training data to a speech recognition component, the training data being in a format recognizable by the speech recognition component and being provided based on the context;

receiving, by the one or more processors, a textual transcript corresponding to the audio file from the speech recognition component;

processing, by the one or more processors, the textual transcript to identify one or more action items by identifying one or more concepts within the textual transcript and matching the one or more concepts to respective transitions in an automaton; and

providing the one or more action items for display to one or more users.

17. A system, comprising:

a computing device; and

a computer-readable storage device coupled to the computing device and having instructions stored thereon which, when executed by the computing device, cause the computing device to perform operations for providing action items from an audio file within an enterprise context, the operations comprising;

determining a context of the audio file that is to be processed based on a user input indicating a training data selection;

providing, training data to a speech recognition component, the training data being in a format recognizable by the speech recognition component and being provided based on the context;

receiving a textual transcript corresponding to the audio file from the speech recognition component;

processing the textual transcript to identify one or more action items by identifying one or more concepts within the textual transcript and matching the one or more concepts to respective transitions in an automaton; and

providing the one or more action items for display to one or more users.