Determining domain salience ranking from ambiguous words in natural speech

Patent metadata

Field	Value
Publication number	US-9734193-B2
Application number	US-201414490372-A
Country	US
Kind code	B2
Filing date	Sep 18, 2014
Priority date	May 30, 2014
Publication date	Aug 15, 2017
Grant date	Aug 15, 2017

Abstract

Systems and processes for identifying relevant domains for user inputs that include one or more ambiguous words are disclosed. The ambiguous words include words that may or may not refer to a named entity, such as a song, movie, book, etc. In one example, a textual representation of user speech can be received and processed to identify a candidate named entity. The possible parts of speech of the candidate named entity can be determined and compared to a predetermined set of parts of speech. In response to determining that the possible parts of speech of the candidate named entity do not include one or more of the predetermined set of parts of speech, a saliency score associated with the candidate named entity can be lowered. A domain for processing the textual representation of user speech can then be identified using the saliency score associated with the candidate named entity.

First claim

What is claimed is:

1. A method for operating a virtual assistant on an electronic device, the method comprising:
receiving a textual representation of user speech;
identifying a candidate named entity from the textual representation of user speech, wherein the candidate named entity is associated with a plurality of saliency scores, each saliency score of the plurality of saliency scores representing a relationship strength between the candidate named entity and a respective domain of a plurality of domains;
determining possible parts of speech of the candidate named entity;
determining whether the possible parts of speech of the candidate named entity comprises one or more of a predetermined set of parts of speech;
in response to determining that the possible parts of speech of the candidate named entity do not comprise one or more of the predetermined set of parts of speech, lowering a saliency score of the plurality of saliency scores associated with the candidate named entity;
identifying a domain of the plurality of domains for processing the textual representation of user speech based at least in part on the lowered saliency score associated with the candidate named entity; and
performing, by the virtual assistant, one or more tasks based on the identified domain to present an output.

2. The method of claim 1 , wherein identifying the candidate named entity from the textual representation of user speech comprises:
identifying a request template based on the textual representation of user speech, wherein the request template comprises a variable entry for a named entity; and
identifying one or more words positioned within the textual representation of user speech at a location corresponding to the variable entry for the named entity, wherein the candidate named entity comprises the one or more words.

3. The method of claim 1 , wherein determining possible parts of speech of the candidate named entity comprises determining all possible parts of speech of the candidate named entity.

4. The method of claim 1 , wherein the candidate named entity comprises a string of two or more words, and wherein determining possible parts of speech of the candidate named entity comprises determining possible parts of speech of a last word of the string of two or more words.

5. The method of claim 1 , wherein determining possible parts of speech of the candidate named entity comprises searching for the candidate named entity in a dictionary to determine the possible parts of speech of the candidate named entity.

6. The method of claim 1 , wherein the predetermined set of parts of speech consists of a noun, a proper noun, and an unknown part of speech.

7. The method of claim 1 , wherein identifying the domain of the plurality of domains for processing the textual representation of user speech is further based on saliency scores associated with other words in the textual representation of user speech.

8. The method of claim 1 , further comprising determining a user intent within the identified domain based on the candidate named entity.

9. The method of claim 1 , wherein lowering the saliency score of the plurality of saliency scores associated with the candidate named entity comprises lowering the saliency score by a predetermined amount.

10. The method of claim 1 , wherein lowering the saliency score of the plurality of saliency scores associated with the candidate named entity comprises lowering the saliency score to a predetermined value.

11. The method of claim 1 , further comprising:
in response to determining that the possible parts of speech of the candidate named entity do not comprise one or more of the predetermined set of parts of speech, lowering each saliency score of the plurality of saliency scores by a predetermined amount.

12. The method of claim 1 , further comprising:
in response to determining that the possible parts of speech of the candidate named entity do not comprise one or more of the predetermined set of parts of speech, lowering each saliency score of the plurality of saliency scores to a predetermined value.

13. A non-transitory computer-readable storage medium comprising instructions for:
receiving a textual representation of user speech;
identifying a candidate named entity from the textual representation of user speech, wherein the candidate named entity is associated with a plurality of saliency scores, each saliency score of the plurality of saliency scores representing a relationship strength between the candidate named entity and a respective domain of a plurality of domains;
determining possible parts of speech of the candidate named entity;
determining whether the possible parts of speech of the candidate named entity comprises one or more of a predetermined set of parts of speech;
in response to determining that the possible parts of speech of the candidate named entity do not comprise one or more of the predetermined set of parts of speech, lowering a saliency score of the plurality of saliency scores associated with the candidate named entity;
identifying a domain of the plurality of domains for processing the textual representation of user speech based at least in part on the lowered saliency score associated with the candidate named entity; and
performing, by a virtual assistant on an electronic device, one or more tasks based on the identified domain to present an output.

14. The non-transitory computer-readable storage medium of claim 13 , wherein identifying the candidate named entity from the textual representation of user speech comprises:
identifying a request template based on the textual representation of user speech, wherein the request template comprises a variable entry for a named entity; and
identifying one or more words positioned within the textual representation of user speech at a location corresponding to the variable entry for the named entity, wherein the candidate named entity comprises the one or more words.

15. The non-transitory computer-readable storage medium of claim 13 , wherein determining possible parts of speech of the candidate named entity comprises determining all possible parts of speech of the candidate named entity.

16. The non-transitory computer-readable storage medium of claim 13 , wherein the candidate named entity comprises a string of two or more words, and wherein determining possible parts of speech of the candidate named entity comprises determining possible parts of speech of a last word of the string of two or more words.

17. The non-transitory computer-readable storage medium of claim 13 , wherein determining possible parts of speech of the candidate named entity comprises searching for the candidate named entity in a dictionary to determine the possible parts of speech of the candidate named entity.

18. A system comprising: an electronic device having: one or more processors; memory; and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for:
receiving a textual representation of user speech;
identifying a candidate named entity from the textual representation of user speech, wherein the candidate named entity is associated with a plurality of saliency scores, each saliency score of the plurality of saliency scores representing a relationship strength between the candidate named entity and a respective domain of a plurality of domains;
determining possible parts of speech of the candidate named entity;
determining whether

Assignees

• Apple Inc

Inventors

• Rhoten George
• Treadgold Nicholas K

Classifications

• G06F16/90332

  Natural language query formulation or dialogue systems · CPC title

• G06F40/205

  Parsing · CPC title

• G06F16/243 Primary

  Natural language query formulation · CPC title

• G06F40/30

  Semantic analysis · CPC title

• G06F40/295

  Named entity recognition · CPC title