Aspect-Level Sentiment Analysis Via Convolution over Dependency Tree

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5679–5688, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5679 Aspect-Level Sentiment Analysis Via Convolution over Dependency Tree Kai Sun BDBC and SKLSDE Beihang University,China sunkai@buaa.edu.cn Richong Zhang∗ BDBC and SKLSDE Beihang University, China zhangrc@act.buaa.edu.cn Samuel Mensah BDBC and SKLSDE Beihang University, China samensah@buaa.edu.cn Yongyi Mao School of EECS University of Ottawa,Canada ymao@uottawa.ca Xudong Liu BDBC and SKLSDE Beihang University, Beijing, China liuxd@act.buaa.edu.cn Abstract We propose a method based on neural net- works to identify the sentiment polarity of opinion words expressed on a specific aspect of a sentence. Although a large majority of works typically focus on leveraging the ex- pressive power of neural networks in handling this task, we explore the possibility of inte- grating dependency trees with neural networks for representation learning. To this end, we present a convolution over a dependency tree (CDT) model which exploits a Bi-directional Long Short Term Memory (Bi-LSTM) to learn representations for features of a sentence, and further enhance the embeddings with a graph convolutional network (GCN) which operates directly on the dependency tree of the sen- tence. Our approach propagates both contex- tual and dependency information from opin- ion words to aspect words, offering discrimi- native properties for supervision. Experimen- tal results ranks our approach as the new state- of-the-art in aspect-based sentiment classifica- tion. 1 Introduction The current explosion in digital technology in re- cent years has led to a vast amount of opinion- ated materials on the internet. In particular, in- dividuals have expressed opinions on several as- pects of products, services, blogs, and comments which are deemed to be influential, especially when making purchase decisions based on prod- uct reviews (Schouten and Frasincar, 2015). How- ever, due to the voluminous amount of content on- line, sifting through reviews to learn knowledge of opinions expressed on specific aspects of a re- view is cumbersome. This fact has led to an in- crease in research in aspect-based sentiment anal- ysis (ABSA), which aims to find scalable solu- tions to address the problem automatically. More ∗Corresponding author specifically, ABSA involves two tasks: (1) to iden- tify aspects of a sentence, and (2) to determine the sentiment polarity (e.g. positive, negative, neutral) expressed on a specific aspect. In this paper, we focus on the second task: aspect-based sentiment classification. We ordered the special grilled branzino , that was so infused with bone, it was difficult to eat. TO NN VBD DT JJ WDT VBD RB VBN IN NN PRP VB VBD JJ PRP JJ TO We ordered the special grilled branzino, that was so infused with bone, it was difficult to eat Figure 1: An example of a dependency tree where an opinion word (blue) and the specific aspect expression (red) are connected with other word tokens based on their syntactic dependencies. With the aim to address the classification task, several methods have been developed. …

Content truncated to 500 words