Pinyin-based semantic identification method and apparatus, and man-machine conversation system

Abstract

The invention provides a pinyin-based semantic identification method and apparatus, and a man-machine conversation system. The semantic identification method comprises the steps of obtaining a pinyinsequence of words in a to-be-identified statement; obtaining word vectors of pinyin fragments in the pinyin sequence by using a word embedding model; combining the word vectors of the pinyin fragmentsin the pinyin sequence into sentence vectors of the to-be-identified statement; inputting the sentence vectors of the to-be-identified statement to a neural network, thereby obtaining output vectorsof the to-be-identified statement; identifying whether the to-be-identified statement is similar to a reference statement in semantics or not by comparing the output vectors of the to-be-identified statement with output vectors of the reference statement; and in response to operation of identifying that the to-be-identified statement is similar to the reference statement in semantics, taking semanteme of the reference statement as semanteme of the to-be-identified statement.

Description

Technical field [0001] The embodiments of the present invention relate to the field of man-machine dialogue, and in particular to a semantic recognition method, device, and man-machine dialogue system. Background technique [0002] With the rapid development of the number of network intelligent devices, the forms of the devices are also various, and the interaction methods are also different. Especially with the rise of voice recognition technology companies, voice recognition technology has become more and more mature, and the scope of applications has also increased. The broader, the voice-based human-computer interaction method is generally regarded as a more natural human-computer interaction method. In the actual application process, the current speech recognition function generally converts the speech signal into text information, but it is difficult to guarantee its accuracy. Sometimes the recognition result returns words with similar pronunciation, but the meaning of the ...

AI Technical Summary

Problems solved by technology

The implementation steps of this method are cumbersome, and the processing of error-prone word pairs requires manual labeling, which further increases the cost

Images