Syntax-Aware Aspect Level Sentiment Classification with Graph Attention Networks

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5469–5477, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5469 Syntax-Aware Aspect Level Sentiment Classification with Graph Attention Networks Binxuan Huang School of Computer Science Carnegie Mellon University binxuanh@cs.cmu.edu Kathleen M. Carley School of Computer Science Carnegie Mellon University kathleen.carley@cs.cmu.edu Abstract Aspect level sentiment classication aims to identify the sentiment expressed towards an aspect given a context sentence. Previous neural network based methods largely ignore the syntax structure in one sentence. In this paper, we propose a novel target-dependent graph attention network (TD-GAT) for aspect level sentiment classification, which explicitly utilizes the dependency relationship among words. Using the dependency graph, it propa- gates sentiment features directly from the syn- tactic context of an aspect target. In our experi- ments, we show our method outperforms mul- tiple baselines with GloVe embeddings. We also demonstrate that using BERT represen- tations further substantially boosts the perfor- mance. 1 Introduction Aspect level sentiment classification aims to iden- tify the sentiment polarity (eg. positive, negative, neutral) of an aspect target in its context sentence. Compared to sentence-level sentiment classifica- tion, which tries to detect the overall sentiment in a sentence, it is a more fine-grained task. Aspect level sentiment classification can distinguish sen- timent polarity for multiple aspects in a sentence with various sentiment polarity, while sentence- level sentiment classification often fails in these conditions (Jiang et al., 2011). For example, in a sentence “great food but the service was dreadful”, the sentiment polarity for aspects “food” and “ser- vice” are positive and negative respectively. In this case, however, it is hard to determine the overall sentiment since the sentence is mixed with posi- tive and negative expressions. Typically, researchers use machine learning al- gorithms to classify the sentiment of given as- pects in sentences. Some early work manually designs features, eg. sentiment lexicons and lin- guistic features, to train classifiers for aspect level sentiment classification (Jiang et al., 2011; Wag- ner et al., 2014). Later, various neural network- based methods became popular for this task (Tang et al., 2016b; Wang et al., 2016), as they do not re- quire manual feature engineering. Most of them are based on long short-term memory (LSTM) neural networks (Tang et al., 2016a; Huang et al., 2018) and few of them use convolutional neural networks (CNN) (Huang and Carley, 2018; Xue and Li, 2018). Most of these neural network based methods treat a sentence as a word sequence and embed aspect information into the sentence representa- tion via various methods, eg. attention (Wang et al., 2016) and gate (Huang and Carley, 2018). These methods largely ignore the syntactic struc- ture of the sentence, which would be beneficial to identify sentiment features directly related to the aspect target. When an aspect term is sepa- rated away from its sentiment phrase, it is hard to find the associated sentiment words in a sequence. For example, …

Content truncated to 500 words