Named entity resolution in spoken language processing

摘要

Features are disclosed for determining an element of a user utterance or user intent in conjunction with one or more related elements of the user utterance or user intent. A user utterance may be transcribed by an automatic speech recognition (“ASR”) module, and the results may be provided to a natural language understanding (“NLU”) module. The NLU module may perform named entity recognition, intent classification, and/or other processes on the ASR results. In addition, the NLU module may determine or verify the values associated with the recognized named entities using a data store of known values. When two or more named entities are related, their values may be determined and/or verified in conjunction with each other in order to preserve the relationship between them.

主权项

1. A system comprising:
a computer-readable memory storing executable instructions; and one or more processors in communication with the computer-readable memory, wherein the one or more processors are programmed by the executable instructions to at least:
receive data regarding an utterance of a user, the utterance associated with a user request;generating first speech recognition results from the utterance and a language processing model;determine, based at least partly on the first speech recognition results, a first named entity and a second named entity;search a data store using the first named entity to determine a first resolved named entity associated with a first confidence score, wherein the first confidence score indicates a likelihood that the first resolved named entity corresponds to the first named entity;search the data store using the second named entity to determine a second resolved named entity associated with a second confidence score, wherein the second confidence score indicates a likelihood that the second resolved named entity corresponds to the second named entity, and wherein the data store comprises information indicating that the first resolved named entity is associated with the second resolved named entity;determine that the first confidence score associated with the first resolved named entity meets or exceeds a threshold confidence score;determine a combined score based on the confidence score associated with the first resolved named entity and the confidence score associated with the second resolved named entity;determine that the combined score does not meet or exceed the threshold confidence score;obtain second speech recognition results from the first speech recognition results and the first resolved named entity;determine, based at least partly on the second speech recognition results, a third resolved named entity included in the data store, wherein the third resolved named entity is associated with a third confidence score that meets or exceeds the threshold confidence score; andgenerate a response based at least partly on the third resolved named entity.