Incorporation of external knowledge in multimodal dialog systems

• US 20070038445A1
• Filed: 05/05/2006
• Published: 02/15/2007
• Est. Priority Date: 05/05/2005
• Status: Active Grant

• Current Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Original Assignee: Nuance Communications, Inc. (Microsoft Corporation)
• Inventors: Helbing, Marc, Reifenrath, Klaus
• Granted Patent: US 7,668,716 B2
• Time in Patent Office: Days
• US Class Current: 704/235
• CPC Class Codes: G10L 15/1815 Semantic context, e.g. disambiguation of the recognition hypotheses based on word meaning
• CPC Class Codes: G10L 15/1822 Parsing for meaning understanding

Abstract

A device improves speech recognition accuracy by utilizing an external knowledge source. The device receives a speech recognition result from an automatic speech recognition (ASR) engine, the speech recognition result including a plurality of ordered interpretations, wherein each of the ordered interpretations includes a plurality of information items. The device analyzes and filters the plurality of interpretations using an external knowledge source to create a filtered plurality of ordered interpretations. The device stores the filtered plurality of ordered interpretations to a memory. The device transmits the filtered plurality of ordered interpretations to a dialog manager module to create a textual output. Alternatively, the dialog manager module retrieves the filtered plurality of ordered interpretations from a memory.

Claims

1. A method of improving speech recognition, comprising:

receiving a speech recognition result from an automatic speech recognition (ASR) engine, the speech recognition result including a plurality of ordered interpretations, wherein each of the ordered interpretations includes a plurality of information items; and

analyzing the plurality of interpretations using an external knowledge source to create a filtered plurality of ordered interpretations.

2. The method of claim 1, further including storing the filtered plurality of ordered interpretations.

3. The method of claim 1, further including transmitting the filtered plurality of ordered interpretations to a dialog manager module to create a textual output.

4. The method of claim 1, wherein the external knowledge source includes a directory with name data and during the analyzing of the plurality of interpretations, an external knowledge processor compares at least one information item in the plurality of information items from each of the plurality of interpretations with the name data directory to determine if any invalid name combinations are present in the at least one information item.

5. The method of claim 4, further including modifying the at least one information item with a valid name combination.

6. The method of claim 4, further including deleting the at least one information item because the at least one informational item was an invalid name combination.

7. The method of claim 1, wherein the external knowledge source includes a calendar with date information and during the analyzing of the plurality of interpretations, an external knowledge processor compares at least one information item in the plurality of information items for each of the plurality of interpretations with the calendar to detect if any invalid date combinations are present in the at least one information item.

8. The method of claim 7, further including modifying the at least one information item with a valid date combination or deleting the at least one information item with an invalid date combination.

9. The method of claim 1, wherein the external knowledge source includes a position location information, and during the analyzing of the plurality of interpretations, an external knowledge processor determines if any of the plurality of information items relate to travel requests, and if so, adds the position location information to each information item of the plurality of information items relating to travel.

10. The method of claim 9, wherein the external knowledge processor further increases a confidence score for each of the plurality of information items relating to travel.

11. The method of claim 1, wherein the external knowledge source includes a geographical database, and during the analyzing of the plurality of interpretations, if the external knowledge processor identifies at least one information item in the plurality of information items is a city name, the external knowledge processor adds a corresponding state name from the geographical database to the identified information item.

12. The method of claim 1, wherein the external knowledge source includes a database having a caller identifier and corresponding preferences and an external knowledge processor determines that an incoming caller identifier has a match in the caller identifier database, then the external knowledge processor retrieves a matching preference for the caller identifier from the database.

13. The method of claim 12, wherein the external knowledge processor generates confidence factors for the plurality of filtered ordered interpretations based on the retrieved matching preference.

14. A method of improving speech recognition, comprising:

receiving a speech recognition result from an automatic speech recognition (ASR) engine, the speech recognition result including a first plurality of ordered interpretations, wherein each of the first plurality of ordered interpretations includes a first plurality of information items; and

analyzing the first plurality of interpretations using an external knowledge source to create a first filtered plurality of ordered interpretations;

receiving a second plurality of ordered interpretations, wherein each of the second plurality of ordered interpretations includes a second plurality of information items;

analyzing the second plurality of ordered interpretations using the external knowledge source and first filter plurality of ordered interpretations to see if any of the second plurality of information items in each of the ordered interpretations can be combined with any information items in the first filtered plurality of interpretations; and

creating a second filtered plurality of ordered interpretations.

15. The method of claim 14, further including storing the second filtered plurality of ordered interpretations.

16. The method of claim 14, further including transmitting the second filtered plurality of

ordered interpretations to a dialog manager module to create a textual output.