Voice endpoint determination

Abstract

Methods, systems, and apparatus are described, including computer programs encoded on computer storage media, for determining endpoints of speech. In one aspect, a method includes an act of accessing voice query log data including voice queries uttered by a particular user. The acts also include determining a pause threshold from the voice query log data including the voice query uttered by the particular user based on the voice query log data including the voice query uttered by the particular user. The actions also include receiving an utterance from the particular user. The actions also include determining that the particular user has stopped speaking for a period of time at least equal to a pause threshold. The acts also include processing the utterance as a voice query based on determining that the particular user has stopped speaking for a period of time at least equal to the pause threshold.

Description

[0001] Cross References to Related Applications

[0002] This application claims the benefit of US Provisional Application No. 62 / 243,463, filed October 19, 2015, the contents of which are incorporated herein by reference. technical field

[0003] The present disclosure relates generally to speech recognition, and one particular implementation relates to endpoint determination for speech. Background technique

[0004] Natural language processing systems typically use endpoints to determine when a user has started and finished speaking. In determining when an utterance begins or ends, some traditional endpointers evaluate the duration of pauses between words. For example, if the user says "what is <long pause> for dinner", a traditional endpoint can segment the speech input at the long pause and can instruct the natural language processing system to try to process Incomplete phrase "whatis (what to eat)" instead of complete phrase "what is for dinner (what to eat for ...

Images