Automatic Language Model Update

• US 20130006640A1
• Filed: 09/14/2012
• Published: 01/03/2013
• Est. Priority Date: 04/03/2006
• Status: Active Grant

• Current Assignee: Google LLC (Alphabet Inc.)
• Original Assignee: Google Inc. (Alphabet Inc.)
• Inventors: Cohen, Michael H., Baluja, Shumeet, Moreno, Pedro J.
• Granted Patent: US 8,447,600 B2
• Time in Patent Office: Days
• US Class Current: 704/257
• CPC Class Codes: G10L 15/06 Creation of reference templates; Training of speech recognition systems, e.g. adaptation to the characteristics of the speaker's voice G10L15/14 takes precedence
• CPC Class Codes: G10L 15/063 Training
• CPC Class Codes: G10L 15/065 Adaptation
• CPC Class Codes: G10L 15/187 Phonemic context, e.g. pronunciation rules, phonotactical constraints or phoneme n-grams
• CPC Class Codes: G10L 15/26 Speech to text systems G10L15/08 takes precedence
• CPC Class Codes: G10L 2015/0635 updating or merging of old and new templates; Mean values; Weighting

Abstract

A method for generating a speech recognition model includes accessing a baseline speech recognition model, obtaining information related to recent language usage from search queries, and modifying the speech recognition model to revise probabilities of a portion of a sound occurrence based on the information. The portion of a sound may include a word. Also, a method for generating a speech recognition model, includes receiving at a search engine from a remote device an audio recording and a transcript that substantially represents at least a portion of the audio recording, synchronizing the transcript with the audio recording, extracting one or more letters from the transcript and extracting the associated pronunciation of the one or more letters from the audio recording, and generating a dictionary entry in a pronunciation dictionary.

Claims

1. A computer-implemented method, comprising:

obtaining, at a server system, information that identifies a level of inclusion of a particular word in queries that were received over a period of time; and

modifying, by the server system, a speech recognition model to revise occurrence data for the particular word based on the level of inclusion of the word in the queries that were received over the period of time.

2. The computer-implemented method of claim 1, wherein the queries that were received over the period of time is a portion of all queries that were received and that include the word.

3. The computer-implemented method of claim 1, wherein revising occurrence data for the particular word comprises weighting a greater prominence for inclusion of the particular word in the queries that were received over the period of time than inclusion of the particular word in queries that were received outside of the period of time.

4. The computer-implemented method of claim 1, wherein the period of time when the queries were received is a period of time of receipt of the queries by a search engine system, and wherein the queries were received from a plurality of computerized devices.

5. The computer-implemented method of claim 1, wherein the queries are textual search queries.

6. The computer-implemented method of claim 1, wherein the queries are verbal search queries.

7. The computer-implemented method of claim 1, further comprising:

receiving, at the server system and from a mobile computing device, a verbal query; and

translating, by the server system and using the modified speech recognition model, the verbal query into a textual query, wherein the textual query includes the particular word.

8. The computer-implemented method of claim 7, further comprising:

collecting search results that are responsive to the textual query by accessing an index of documents in a search engine system; and

transmitting the collected search results for use by the mobile computing device.

9. The computer-implemented method of claim 8, further comprising transmitting, with the collected search results and for use by the mobile computing device, the textual query.

10. The computer-implemented method of claim 8, wherein the collected search results and textual query are formatted for transmission so as to cause the mobile computing device to automatically display, in a concurrent view, the collected search results and the textual query.

11. The computer-implemented method of claim 1, further comprising modifying, by the server system, the speech recognition model to revise a concurrence weighting for the particular word, wherein the concurrence weighting provides a determination of the likelihood of appearance of the particular word and a second word being adjacent, and wherein the concurrence weighting is generated based on the presence of the particular word and the second word together in one or more search queries.

12. The computer-implemented method of claim 11, wherein the concurrence weighting is generated based on the particular word and the second word being adjacent in one or more search queries.

13. A system comprising:

a server system programmed to obtain information that identifies a level of inclusion of a particular word in queries that were received over a defined period of time; and

an updater to modify a speech recognition model to revise occurrence data for the particular word based on the level of inclusion of the word in queries that were received over the defined period of time.

14. The system of claim 13, wherein the updater is stored on a mobile computing device.

15. The system of claim 14, wherein the server system transmits information to the mobile computing device that enables the speech recognition model to be modified by the updater.

16. A system comprising:

a server system programmed to receive information that identifies when queries that include a particular word were received; and

means for modifying a speech recognition model to revise occurrence data for the particular word based on when the particular word was received in queries.

17. The system of claim 16, further comprising a search engine system to receive the queries as search requests.