Situation aware NLU/NLP
First Claim
Patent Images
1. A system for natural language processing comprising:
- a computing device that receives natural language input from a user;
a data store storing a user state model comprising user context data that characterizes an operating context of the computing device;
memory storing instructions that, when executed by a processor of the computing device, cause the system to (a) perform a natural language understanding (NLU) process that produces a final interpretation of the natural language input and (b) bias performance of the NLU process using the user context data of the user state model;
wherein the NLU process comprises;
processing the natural language input in parallel using a plurality of domain pipelines to produce, by each of the plurality of domain pipelines, one or more domain interpretation candidates, each domain pipeline representing a different subject domain of related concepts,ranking each of the domain interpretation candidates produced by the plurality of domain pipelines, andselecting, based on the ranking, one of the domain interpretation candidates produced by the plurality of domain pipelines as the final interpretation of the natural language input; and
wherein each domain pipeline comprises a plurality of stages, the plurality of stages comprising;
a mention detection stage that assigns, to one or more words of the natural language input, a tag indicating a semantic concept of the subject domain represented by the domain pipeline,an interpretation stage that produces the one or more domain interpretation candidates, each domain interpretation candidate having a confidence value and the one or more domain interpretation candidates being ranked according to a confidence value for each domain interpretation candidate,a query intent classification stage that determines a query intent for the natural language input, andan evidence ranking stage that re-ranks the one or more domain interpretation candidates using a classifier trained on data representing the query intent.
2 Assignments
0 Petitions
Accused Products
Abstract
An arrangement and corresponding method are described for natural language processing. A natural language understanding (NLU) arrangement processes a natural language input to determine a corresponding sentence-level interpretation. A user state component maintains user context data that characterizes an operating context of the NLU arrangement. Operation of the NLU arrangement is biased by the user context data.
41 Citations
20 Claims
-
1. A system for natural language processing comprising:
-
a computing device that receives natural language input from a user; a data store storing a user state model comprising user context data that characterizes an operating context of the computing device; memory storing instructions that, when executed by a processor of the computing device, cause the system to (a) perform a natural language understanding (NLU) process that produces a final interpretation of the natural language input and (b) bias performance of the NLU process using the user context data of the user state model; wherein the NLU process comprises; processing the natural language input in parallel using a plurality of domain pipelines to produce, by each of the plurality of domain pipelines, one or more domain interpretation candidates, each domain pipeline representing a different subject domain of related concepts, ranking each of the domain interpretation candidates produced by the plurality of domain pipelines, and selecting, based on the ranking, one of the domain interpretation candidates produced by the plurality of domain pipelines as the final interpretation of the natural language input; and wherein each domain pipeline comprises a plurality of stages, the plurality of stages comprising; a mention detection stage that assigns, to one or more words of the natural language input, a tag indicating a semantic concept of the subject domain represented by the domain pipeline, an interpretation stage that produces the one or more domain interpretation candidates, each domain interpretation candidate having a confidence value and the one or more domain interpretation candidates being ranked according to a confidence value for each domain interpretation candidate, a query intent classification stage that determines a query intent for the natural language input, and an evidence ranking stage that re-ranks the one or more domain interpretation candidates using a classifier trained on data representing the query intent. - View Dependent Claims (2, 3, 4, 5, 11, 13, 14, 15, 16)
-
-
6. A method for natural language processing comprising:
-
storing, at a data store of a computing device, a user state model comprising user context data that characterizes an operating context of the computing device; receiving, at the computing device, natural language input from the user; performing, by a processor of the computing device, a natural language understanding (NLU) process that produces a final interpretation of the natural language input; wherein performance of the NLU process is biased using the user context data of the user state model; wherein the NLU process comprises; processing the natural language input in parallel using a plurality of domain pipelines to produce, by each of the plurality of domain pipelines, one or more domain interpretation candidates, each domain pipeline representing a different subject domain of related concepts, ranking each of the domain interpretation candidates produced by the plurality of domain pipelines, and selecting, based on the ranking, one of the domain interpretation candidates produced by the plurality of domain pipelines as the final interpretation of the natural language input; and wherein each domain pipeline comprises a plurality of stages, the plurality of stages comprising; a mention detection stage that assigns, to one or more words of the natural language input, a tag indicating a semantic concept of the subject domain represented by the domain pipeline, an interpretation stage that produces the one or more domain interpretation candidates, each domain interpretation candidate having a confidence value and the one or more domain interpretation candidates being ranked according to a confidence value for each domain interpretation candidate, a query intent classification stage that determines a query intent for the natural language input, and an evidence ranking stage that re-ranks the one or more domain interpretation candidates using a classifier trained on data representing the query intent. - View Dependent Claims (7, 8, 9, 10, 12, 17, 18, 19, 20)
-
Specification