Open AccessChinese Reviews Generation Based on HM-BiLSTM Model
Author(s) -
Jianglin Yuan,
Zhigang Guo,
Gang Chen,
Yihe Sun,
Ruipeng Yang
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1325/1/012075
Subject(s) - computer science , subject (documents) , sentence , artificial intelligence , theme (computing) , imitation , natural language processing , mechanism (biology) , expression (computer science) , linguistics , psychology , world wide web , social psychology , philosophy , epistemology , programming language
HM-BiLSTM model is proposed to solve the problem of inaccurate expression of Chinese comment with specific topic, which is generated by existing deep learning models. Firstly, HM-BiLSTM model was created to generate comments with subject attention mechanism algorithm whose purpose was to assist to constrain the theme. Then, imitation writing sentence was retrieved to extract syntactic structure information from Chinese reviews corpus. The subject as well as structure information was both encoded. Lastly, Chinese comments were generated recurrently by HM-BiLSTM model under the constraints of the subject attention mechanism algorithm and syntactic structure information. The experiment results showed that the model generated comments with better quality and accurate theme.