End-to-end aspect category sentiment analysis based on type graph convolutional networks

For the existing aspect category sentiment analysis research,most of the aspects are given for sentiment extraction,and this pipeline method is prone to error accumulation,and the use of graph convolutional neural network for aspect category sentiment analysis does not fully utilize the dependency type information between words,so it cannot enhance feature extraction.This paper proposes an end-to-end aspect category sentiment analysis(ETESA)model based on type graph convolutional networks.The model uses the bidirectional encoder representation from transformers(BERT)pretraining model to obtain aspect categories and word vectors containing contextual dynamic semantic information,which can solve the problem of polysemy;when using graph convolutional network(GCN)for feature extraction,the fusion operation of word vectors and initialization tensor of dependency types can obtain the importance values of different dependency types and enhance the text feature representation;by transforming aspect category and sentiment pair extraction into multiple single-label classification problems,aspect category and sentiment can be extracted simultaneously in an end-to-end way and solve the problem of error accumulation.Experiments are tested on three public datasets,and the results show that the ETESA model can achieve higher Precision,Recall and F1 value,proving the effectiveness of the model.

Graph Enhanced Transformer for Aspect Category Detection

Aspect category detection is one challenging subtask of aspect based sentiment analysis, which categorizes a review sentence into a set of predefined aspect categories. Most existing methods regard the aspect category detection as a flat classification problem. However, aspect categories are inter-related, and they are usually organized with a hierarchical tree structure. To leverage the structure information, this paper proposes a hierarchical multi-label classification model to detect aspect categories and uses a graph enhanced transformer network to integrate label dependency information into prediction features. Experiments have been conducted on four widely-used benchmark datasets, showing that the proposed model outperforms all strong baselines.