Intention recognition method based on deep learning network

Abstract

The invention relates to the field of intelligent recognition and discloses an intention recognition method based on a deep learning network. The technical problem that the intention recognition accuracy is not high is solved. According to the technical scheme, the method is characterized in that features of a first deep learning network are migrated to a second deep learning network, and the method mainly comprises the steps that data sets in all fields are converted into word sequences WS and corresponding pinyin sequences PS, and meanwhile the data sets in a certain field are manually labeled and converted into word sequences WD, pinyin sequences PD and labels. Inputting the word sequence WS and the pinyin sequence PS into a first deep learning network for training to obtain a languagemodel, a coding layer parameter matrix of the language model is initialized and updated, then a word sequence WD and a Pinyin sequence PD are input into a second deep learning network for coding, weighting is input into the second deep learning network for training an intention recognition model, and the intention recognition model is high in intention recognition accuracy.

Description

technical field [0001] The present disclosure relates to the field of intelligent identification, and in particular to an intention identification method based on a deep learning network. Background technique [0002] In the field of human-computer dialogue, intent recognition is one of the core technologies, and the understanding of natural semantics is one of the prerequisites for the realization of human-computer dialogue. At present, the basic methods of intent recognition include the rule method based on dictionaries and templates, the user’s intent based on query click logs, and classification models. The industry agrees that the biggest difficulty in intent recognition lies in the acquisition of labeled data. Research methods with a huge amount of data such as Pinyin sequences are rarely carried out, and the recognition model trained by character and Pinyin feature vectors is almost in a blank state. Contents of the invention [0003] The purpose of the present dis...

AI Technical Summary

Problems solved by technology

At present, the basic methods of intent recognition include the rule method based on dictionaries and templates, the user’s intent based on query click logs, and classification models. The industry agrees that the biggest difficulty in intent recognition lies in the acquisition of labeled data. Research methods with huge amounts of data such as pinyin sequences are rarely carried out, and the recognition model trained by character and pinyin feature vectors is almost blank

Images