Learning to Detect Opinion Snippet for Aspect-Based Sentiment Analysis

Proceedings of the 23rd Conference on Computational Natural Language Learning, pages 970–979 Hong Kong, China, November 3-4, 2019. c⃝2019 Association for Computational Linguistics 970 Learning to Detect Opinion Snippet for Aspect-Based Sentiment Analysis Mengting Hu1∗ Shiwan Zhao2† Honglei Guo2 Renhong Cheng1 Zhong Su2 1 Nankai University 2 IBM Research - China mthu@mail.nankai.edu.cn, {zhaosw, guohl}@cn.ibm.com chengrh@nankai.edu.cn, suzhong@cn.ibm.com Abstract Aspect-based sentiment analysis (ABSA) is to predict the sentiment polarity towards a partic- ular aspect in a sentence. Recently, this task has been widely addressed by the neural at- tention mechanism, which computes attention weights to softly select words for generating aspect-specific sentence representations. The attention is expected to concentrate on opin- ion words for accurate sentiment prediction. However, attention is prone to be distracted by noisy or misleading words, or opinion words from other aspects. In this paper, we pro- pose an alternative hard-selection approach, which determines the start and end positions of the opinion snippet, and selects the words between these two positions for sentiment pre- diction. Specifically, we learn deep associa- tions between the sentence and aspect, and the long-term dependencies within the sentence by leveraging the pre-trained BERT model. We further detect the opinion snippet by self- critical reinforcement learning. Especially, ex- perimental results demonstrate the effective- ness of our method and prove that our hard- selection approach outperforms soft-selection approaches when handling multi-aspect sen- tences. 1 Introduction Aspect-based sentiment analysis (Pang and Lee, 2008; Liu, 2012) is a fine-grained sentiment anal- ysis task which has gained much attention from research and industries. It aims at predicting the sentiment polarity of a particular aspect of the text. With the rapid development of deep learning, this task has been widely addressed by attention- based neural networks (Wang et al., 2016; Ma et al., 2017; Cheng et al., 2017; Tay et al., 2018; ∗ Work performed while interning at IBM Research - China. †Corresponding author. 0.20 0.12 0.09 0.1 0.04 0.05 0.04 Aspect: place Label: Negative Prediction: Positive the food is usually good but it certainly is not a relaxing place to go Figure 1: Example of attention visualization. The at- tention weights of the aspect place are from the model ATAE-LSTM (Wang et al., 2016), a typical attention mechanism used for soft-selection. Wang et al., 2018a). To name a few, Wang et al. (2016) learn to attend on different parts of the sentence given different aspects, then generates aspect-specific sentence representations for sen- timent prediction. Tay et al. (2018) learn to at- tend on correct words based on associative rela- tionships between sentence words and a given as- pect. These attention-based methods have brought the ABSA task remarkable performance improve- ment. Previous attention-based methods can be cate- gorized as soft-selection approaches since the at- tention weights scatter across the whole sentence and every word is taken into consideration with different weights. This usually results in atten- tion distraction (Li et al., 2018b), i.e., attending on noisy or misleading words, or opinion words from other aspects. Take Figure 1 as an …

Content truncated to 500 words