Semantic pre-processing of natural language input in a virtual personal assistant

Patent metadata

Field	Value
Publication number	US-10579835-B1
Application number	US-201414253503-A
Country	US
Kind code	B1
Filing date	Apr 15, 2014
Priority date	May 22, 2013
Publication date	Mar 3, 2020
Grant date	Mar 3, 2020

Abstract

A computing system is operable as a virtual personal assistant (VPA). Among other things, a semantic preprocessor of the VPA recognizes natural language user input that is intended as a reference to a known semantic type, and creates a semantic interpretation of the natural language input. The semantic interpretation may be used by a natural language understanding module of the VPA.

First claim

The invention claimed is:

1. A method for a computing device to pre-process natural language speech captured by an audio input device to improve accuracy of a parser, the method comprising:
with syntactic parser computer instructions, identifying a plurality of syntactic units in an instance of natural language input;
with unit-level semantic parser computer instructions, outputting unit-level semantic interpretations each comprising a canonical representation of a particular subset of the plurality of syntactic units and a data value associated with the canonical representation of the particular subset of the plurality of syntactic units;
with sentence-level semantic parser computer instructions, determining one of the plurality of unit-level semantic interpretations output by the unit-level semantic parser computer instructions to include in a sentence-level semantic interpretation of the instance of natural language input;
with reasoner computer instructions, using the sentence-level semantic interpretation of the instance of natural language input to determine an action to be executed by the computing device;
causing the computing device to execute the action.

2. The method of claim 1 , comprising using the data structure outputted by the unit-level semantic parser computer instructions as input to another unit-level semantic parser.

3. The method of claim 1 , comprising using a second data structure outputted by second unit-level semantic parser computer instructions as input to the sentence-level semantic parser to determine a second semantic interpretation of the same instance of natural language input.

4. The method of claim 1 , comprising executing the syntactic parser computer instructions to verify a data structure outputted by the unit-level semantic parser computer instructions.

5. The method of claim 1 , comprising concurrently identifying a plurality of candidate canonical representations of syntactic units of the same instance of natural language input and using at least one of the candidate canonical representations to create the data structure.

6. The method of claim 1 , comprising iteratively identifying a plurality of candidate canonical representations of syntactic units of the same instance of natural language input and using at least one of the candidate canonical representations to iteratively build the data structure.

7. The method of claim 1 , wherein the syntactic unit includes a disfluency or a pause.

8. The method of claim 1 , wherein the canonical representation of the syntactic unit comprises a type of currency and the data value comprises an amount of the currency.

9. The method of claim 1 , wherein the canonical representation of the syntactic unit comprises a type of account number and the data value comprises an account number.

10. The method of claim 1 , wherein the data value is extracted from the same instance of natural language input or derived from the same instance of natural language input by executing a computer operation.

11. The method of claim 1 , wherein the unit-level semantic parser interfaces with a user-specific model to identify the canonical representation of the syntactic unit.

12. The method of claim 1 , comprising using the data structure as a reference to a portion of the same instance of natural language input.

13. The method of claim 1 , comprising using the semantic interpretation to personalize a natural language dialog involving the computing device and a person interacting with the computing device.

14. The method of claim 1 , comprising using the semantic interpretation to create an intent for a virtual personal assistant computer system.

15. A system for enabling a computing device to pre-process natural language speech captured by an audio input device to improve accuracy of a parser, the method comprising:
with syntactic parser computer instructions, identifying a plurality of syntactic units in an instance of natural language input;
with unit-level semantic parser computer instructions, outputting unit-level semantic interpretations each comprising a canonical representation of a particular subset of the plurality of syntactic units and a data value associated with the canonical representation of the particular subset of the plurality of syntactic units;
with sentence-level semantic parser computer instructions, determining one of the plurality of unit-level semantic interpretations output by the unit-level semantic parser computer instructions to include in a sentence-level semantic interpretation of the instance of natural language input;
with reasoner computer instructions, using the sentence-level semantic interpretation of the instance of natural language input to determine an action to be executed by the computing device;
causing the computing device to execute the action.

16. The system of claim 15 , wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform operations comprising using the data structure outputted by the unit-level semantic parser computer instructions as input to another unit-level semantic parser.

17. The system of claim 15 , wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform operations comprising using a second data structure outputted by second unit-level semantic parser computer instructions as input to the sentence-level semantic parser to determine a second semantic interpretation of the same instance of natural language input.

18. The system of claim 15 , comprising wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform operations comprising executing the syntactic parser computer instructions to verify a data structure outputted by the unit-level semantic parser computer instructions.

19. The system of claim 15 , wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform operations comprising concurrently identifying a plurality of candidate canonical representations of syntactic units of the same instance of natural language input and using at least one of the candidate canonical representations to create the data structure.

20. The system of claim 15 , wherein the instructions, when executed by the one or more processors, cause the one or more processors to perform operations comprising iteratively identifying a plurality of candidate canonical representations of syntactic units of the same instance of natural language input and using at least one of the candidate canonical representations to iteratively build the data structure.

21. The system of claim 15 , wherein the syntactic unit includes a disfluency or a pause.

22. The system of claim 15 , wherein the canonical representation of the syntactic unit comprises a type of currency and the data value comprises an amount of the currency.

23. The system of claim 15 , wherein the canonical representation of the syntactic unit comprises a type of account number and the data value comprises an account number.

24. The system of claim 15 , wherein the data value is extracted from the same instance of natural language input or derived from the same instance of natural language input by executing a computer operation.

25. The system of claim 15 , wherein the unit-level semantic parser interfaces with a user-specific model to identify the canoni

Assignees

• Stanford Res Inst Int

Inventors

• Phillips Aaron B
• Ayan Necip F

Classifications

• G06F40/211 Primary

  Syntactic parsing, e.g. based on context-free grammar [CFG] or unification grammars · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F40/40 Primary

  Processing or translation of natural language (natural language analysis G06F40/20; semantic analysis G06F40/30) · CPC title