Author(s)

Yanli Liu, Liancheng Xu

Corresponding Author

Yanli Liu, Liancheng Xu

ABSTRACT

This paper proposes a dictionary classification method that incorporates hard attention mechanism in the classification of intention domain, an important link in human-computer dialogue.Based on the common method of classification of sentences, through the establishment of intent on the classification of the data of artificial domain dictionary, and then manually collected sentence to clean, cut and remove the stop operation, then iterative matching and dictionary before joining attention mechanism, will be more important information in the sentence first match, through the test on real data sets, this article selects the traditional machine learning method and the emerging neural network as a comparison method, verified: in this paper, on the accuracy and speed is better than the other.

KEYWORDS

Intention recognition, hard attention, short text classification, domain dictionary