Open AccessMulti-task Learning Network based on Attention for Aspect-Based Sentiment Analysis
Author(s) -
Dong Yang,
Jing Wang,
Junwei Wang
Publication year - 2021
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 - 10.1088/1742-6596/1827/1/012173
Subject(s) - computer science , sentiment analysis , pipeline (software) , artificial intelligence , task (project management) , benchmark (surveying) , machine learning , transfer of learning , context (archaeology) , multi task learning , natural language processing , paleontology , management , geodesy , economics , biology , programming language , geography
Aspect-based sentiment analysis is to analyse the sentiment polarity of the aspect in a specific context. The previous methods mostly use the working mode of pipeline model, and internally use Recurrent Neural Network to predict the sentiment polarity of the aspect. This way clearly defines the order of aspect extraction and aspect-level sentiment classification, which will lead to the problem of error transmission. And traditional models have great limitations. In this paper, we propose a multi-task learning network based on attention (MNA), which uses shared word embedding to transmit to downstream tasks, allowing extraction and classification tasks to be processed simultaneously. We design a network architecture based on attention mechanism, and use the improved multi-head attention mechanism to transfer information from extracted tasks to classified tasks. Experimental results on three benchmark datasets show the effectiveness of MNA model.