System and Method for Processing Speech

• US 20100185443A1
• Filed: 03/31/2010
• Published: 07/22/2010
• Est. Priority Date: 12/06/2004
• Status: Active Grant

• Current Assignee: Interactions, LLC
• Original Assignee: AT&T Intellectual Property I LP (AT&T, Inc.)
• Inventors: Knott, Benjamin Anthony, Bushey, Robert R., Korth, Sarah, Martin, John Mills
• Granted Patent: US 8,306,192 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/00 Speech recognition G10L17/00 takes precedence
• CPC Class Codes: H04M 2201/40 using speech recognition speech recognition per se G10L15/00
• CPC Class Codes: H04M 2203/2011 Service processing based on information specified by a party before or during a call, e.g. information, tone or routing selection
• CPC Class Codes: H04M 3/5166 in combination with interactive voice response systems or voice portals, e.g. as front-ends

Abstract

Systems and methods for processing speech are provided. A system may include a speech recognition interface and a processor. The processor may convert speech received from a call at the speech recognition interface to at least one word string. The processor may parse each word string of the at least one word string into first objects and first actions. The processor may access a synonym table to determine second objects and second actions based on the first objects and the first actions. The processor may also select a preferred object and a preferred action from the second objects and the second actions.

Claims

1. A speech recognition system, comprising:

a speech recognition interface; and

a processor coupled to the speech recognition interface to;

convert speech received from a caller at the speech recognition interface to at least one word string;

parse each word string of the at least one word string into first objects and first actions;

access a synonym table to determine second objects and second actions based on the first objects and the first actions; and

select a preferred object and a preferred action from the second objects and the second actions.

2. The speech recognition system of claim 1, wherein the processor is configured to route the caller to a destination based on the preferred action and the preferred object.

3. The speech recognition system of claim 2, wherein the destination is determined from an action-object routing table based on the preferred action and the preferred object.

4. The speech recognition system of claim 1, wherein the processor converts speech received from the caller at the speech recognition interface to the at least one word string based on phonemes.

5. The speech recognition system of claim 1, wherein the processor is configured to assign a probability value to each word string of the at least one word string based at least in part on the coherency of the word string.

6. The speech recognition system of claim 5, wherein the processor creates first action-object pairs from the first actions and the first objects of each word string of the at least one word string, wherein the processor is configured to assign a confidence value to each first action-object pair, and wherein the confidence value for a particular first action-object pair is based at least in part on the probability value of the word string used to create the particular first action-object pair.

7. The speech recognition system of claim 6, wherein the processor creates a second action-object pair from each first action-object pair when the processor accesses the synonym table to determine the second objects and the second actions, wherein the processor is configured to assign a confidence level to each second action-object pair, wherein a particular confidence level for a particular second action-object pair is based at least in part on the confidence value of the first action-object pair used to obtain the particular second action-object pair.

8. The speech recognition system of claim 7, wherein the preferred object and the preferred action are selected from the second action-object pairs based at least in part on the preferred object and the preferred action having a high confidence level compared to other second action-object pairs.

9. The speech recognition system of claim 1, wherein the first objects correspond in part to nouns in the at least one word string.

10. The speech recognition system of claim 1, wherein the first actions correspond in part to verbs in the at least one word string.

11. A computerized method of processing speech, the method comprising:

determining a plurality of objects based on speech input;

determining a plurality of actions based on the speech input;

comparing the objects and the actions with entries in a synonym table to determine synonym objects and synonym actions;

selecting a preferred object and a preferred action from the synonym objects and the synonym actions; and

routing a call that provided the speech input to a destination based on the preferred object and the preferred action.

12. The computerized method of claim 11, wherein determining the plurality of objects and the plurality of actions comprises:

converting the speech input to word strings based on the speech input; and

parsing the word strings into the plurality of objects and the plurality of actions.

13. The computerized method of claim 11, wherein the preferred object and the preferred action are selected at least in part based on confidence levels of the synonym objects and the synonym actions.

14. The computerized method of claim 11, wherein the destination comprises one of a billing department, a balance information service, a new services department, and a technical support department.

15. The computerized method of claim 11, wherein the preferred object is not phonetically related to content of the speech input.

16. The computerized method of claim 11, wherein a particular object determined based on the speech input comprises a natural language phrase and the preferred object associated with the particular object comprises a term related to the natural language phrase.

17. The computerized method of claim 11, wherein a particular action determined based on the speech input comprises a natural language phrase and the preferred action associated with the particular action comprises a term related to the natural language phrase.

18. A computerized method, comprising:

transforming speech input from a call into a plurality of word strings;

converting the word strings into pairs of objects and actions;

determining from a synonym table synonym pairs based on the pairs of objects and actions; and

selecting a preferred pair from the synonym pairs.

19. The computerized method of claim 18, further comprising determining a destination for the call based on an object and an action of the preferred pair.

20. The computerized method of claim 18, wherein selecting the preferred pair comprises choosing a synonym pair based at least in part on a confidence value of the synonym pair.