Methods and system for capturing voice files and rendering them searchable by keyword or phrase

• US 8,731,919 B2
• Filed: 10/16/2008
• Issued: 05/20/2014
• Est. Priority Date: 10/16/2007
• Status: Active Grant

• Current Assignee: Astute, Inc. (Astute Holdings, Inc.)
• Original Assignee: Astute, Inc. (Astute Holdings, Inc.)
• Inventors: George, Alex Kiran
• Primary Examiner(s): He, Jialong
• Application Number:
  US12/288,261
• Publication Number:
  US 20090099845A1
• Time in Patent Office: 2,042 Days
• Field of Search: 704/251, 704/270, 704/235
• US Class Current: 704/235
• CPC Class Codes: G10L 15/183 using context dependencies, e.g. language models
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence

Abstract

A system for capturing voice files and rendering them searchable, comprising one or more devices capable of capturing audio speech electronically, a recorder coupled to the devices for retrieving audio speech, a controller coupled to the recorder, a recognition engine adapted to transcribe audio speech into text, and a database system is disclosed. In the system, the controller causes the recorder to capture audio speech from at least one of the devices, the recorder stores the audio speech as data in the database system, and the recognition engine subsequently retrieves the audio speech data, transcribes the audio speech data into text, and stores the text and data associating the text data with at least the audio speech data in the database system for subsequent retrieval by a search application.

Claims

1. A system for capturing voice files and rendering them searchable, comprising:

(a) a database system having a plurality of grammars stored therein;

(b) at least one device that electronically captures audio speech for a conversation between two or more participants;

(c) a recorder coupled to said at least one device, the recorder capturing audio speech from the device for storage as audio speech data in said database system; and

(d) a speech recognition engine adapted to

transcribe the audio speech data into machine-readable text data in a plurality of transcription passes using grammars selected from said plurality of stored grammars, and

store the machine-readable text data as well as data associating the machine-readable text data with the corresponding audio speech data in the database system for subsequent retrieval by a search application;

wherein the speech recognition engine is adapted to select a grammar from said database system prior to performing a first transcription pass, the grammar for a first transcription pass selected on the basis of

information pertaining to the subject matter or purpose of the conversation, and

information pertaining to one or more of the participants,

and further wherein the recognition engine is adapted to revise the machine-readable text data for the conversation by performing a subsequent transcription pass on the audio speech data using a grammar which was not used in the first transcription pass.

2. The system of claim 1 further comprising:

a controller coupled to the recorder, the controller causing the recorder to capture audio speech from said at least one device.

3. The system of claim 1 wherein the system is adapted to index the machine-readable text data, the index searchable and selected from the group consisting of subject matter, meeting type, conference purpose, and participant personal identification.

4. The system of claim 1, wherein the speech recognition engine is adapted to receive information regarding the subject matter or purpose of the conversation prior to the conversation taking place.

5. The system of claim 4, further comprising a scheduler for scheduling a telephone conversation between two or more participants and providing the information regarding the subject matter or purpose of the conversation to the speech recognition engine for use in selecting the grammar for the first transcription pass.

6. The system of claim 1, wherein the system is adapted to determine the identities of participants, and further wherein the speech recognition engine is adapted to select the grammar for the first transcription pass on the basis of the determined identity of one or more of the participants.

7. The system of claim 6, wherein the system is adapted to determine the identities of participants based on stored voice samples.

8. The system of claim 1, wherein said at least one device for capturing audio is adapted to electronically capture audio speech from a telephone conference between two or more participants, said system further comprising a scheduler adapted to schedule such a telephone conference, provide telephone access information to participants, and provide information regarding the subject matter of a scheduled telephone conference to the speech recognition engine.

9. The system of claim 1, wherein the speech recognition engine is adapted to select a plurality of grammars from said database system prior to performing a first transcription pass, the plurality of grammars selected on the basis of:

(a) received information pertaining to the subject matter or purpose of the conversation, and

(b) received information pertaining to one or more of the participants.

10. The system of claim 1, wherein said system is adapted to store audio speech data in the database system as a plurality of files, with each file associated with a different participant.

11. The system of claim 1, wherein the grammar used for the subsequent transcription pass is selected based on the content of the machine-readable text data resulting from the first transcription pass.

12. A computer-implemented method for capturing voice files and rendering them searchable, comprising the steps of:

(a) recording audio speech data for a conversation between two or more participants, said audio speech data obtained from at least one audio-capable device;

(b) storing the audio speech data in a database system;

(c) selecting and loading into a speech recognition engine a grammar selected from a plurality of stored grammars, wherein said grammar is selected prior to the transcribing step and is selected on the basis of

information pertaining to the subject matter or purpose of the conversation, and

the identity of one or more of the participants;

(d) transcribing the audio speech data into machine-readable text data using the speech recognition engine employing said grammar;

(e) creating at least one data element associating the machine-readable text data with the corresponding audio speech data;

(f) storing the machine-readable text data and the associated data element in a searchable database; and

(f) revising the machine-readable text data by performing a subsequent transcription pass on the audio speech data using another grammar which is different than the previously selected grammar.

13. The method of claim 12, wherein said conversation is a telephone conference, and further comprising the steps of identifying and tracking voice input in order to determine speaker identities.

14. The method of claim 12, wherein said conversation is a scheduled telephone conference, and further comprising the step of receiving from an organizer of the telephone conference information regarding the subject matter of the telephone conference such that the grammar selected prior to the transcribing step is selected on the basis of said received information regarding the subject matter of the telephone conference.

15. The method of claim 14, further comprising the step of determining the identities of telephone conference participants.

16. The method of claim 14, further comprising the step of determining the identities of telephone conference participants based on stored voice samples.

17. The method of claim 12, wherein said audio speech data is stored in the database system as a plurality of files, with each file associated with a different participant.