Transfer Capsule Network for Aspect Level Sentiment Classification

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics, pages 547–556 Florence, Italy, July 28 - August 2, 2019. c⃝2019 Association for Computational Linguistics 547 Transfer Capsule Network for Aspect Level Sentiment Classification Zhuang Chen, Tieyun Qian∗ School of Computer Science, Wuhan University, China {zhchen18, qty}@whu.edu.cn Abstract Aspect-level sentiment classification aims to determine the sentiment polarity of a sentence towards an aspect. Due to the high cost in an- notation, the lack of aspect-level labeled data becomes a major obstacle in this area. On the other hand, document-level labeled data like reviews are easily accessible from online web- sites. These reviews encode sentiment knowl- edge in abundant contexts. In this paper, we propose a Transfer Capsule Network (Tran- sCap) model for transferring document-level knowledge to aspect-level sentiment classi- fication. To this end, we first develop an aspect routing approach to encapsulate the sentence-level semantic representations into semantic capsules from both aspect-level and document-level data. We then extend the dy- namic routing approach to adaptively couple the semantic capsules with the class capsules under the transfer learning framework. Exper- iments on SemEval datasets demonstrate the effectiveness of TransCap. 1 Introduction Aspect-level sentiment classification (ASC) is a fine-grained subtask in sentiment analysis. Given a sentence and an aspect occurring in the sentence, ASC aims to determine the sentiment polarity of the aspect. Traditional methods mostly use ma- chine learning models with handcrafted features to build sentiment classifiers for ASC tasks (Jiang et al., 2011; Mohammad et al., 2013). Such meth- ods need either laborious feature engineering or massive linguistic resources. With the develop- ment of deep learning technique, a number of neural models have been proposed (Wang et al., 2016b; Tang et al., 2016; Chen et al., 2017) for ASC tasks. All these models train classifiers in a supervised manner and require sufficient num- *Corresponding author. ber of labeled data to get promising results. How- ever, the annotation of opinion targets in ASC is extremely expensive. The lack of labeled data is a major obstacle in this field. Publicly available datasets for ASC of- ten contain limited number of training samples. On the other hand, document-level labeled data like reviews are easily accessible from online web- sites such as Yelp and Amazon. Since each re- view has an accompanying rating score indicating user’s overall satisfaction towards an item, such a score can naturally serve as the label of sentiment polarity of the review document. Intuitively, the document-level data contain use- ful sentiment knowledge for analysis on aspect- level data since they may share many linguistic and semantic patterns. Unfortunately, for ASC tasks, only one study (He et al., 2018) has taken the utilization of document-level data into ac- count. The PRET+MULT framework proposed in (He et al., 2018) is a successful attempt by adopting pre-training and multi-task learning ap- proaches. However, their model only shares shal- low embedding and LSTM layers between ASC and DSC (document-level sentiment classifica- tion) tasks. In other words, the document-level knowledge is merely used …

Content truncated to 500 words