• PDF / 827,726 Bytes
• 15 Pages / 439.37 x 666.142 pts Page_size
• 52 Downloads / 245 Views

DOWNLOAD

REPORT

---

A relative position attention network for aspect-based sentiment analysis Chao Wu1,2 · Qingyu Xiong1,2 Kaige Wang1,2

· Min Gao1,2 · Qiude Li1,2 · Yang Yu1,2 ·

Received: 26 July 2019 / Revised: 9 September 2020 / Accepted: 12 September 2020 © Springer-Verlag London Ltd., part of Springer Nature 2020

Abstract Aspect-based sentiment analysis can predict the sentiment polarity of specific aspect terms in the text. Compared to general sentiment analysis, it extracts more useful information and analyzes the sentiment more accurately in the comment text. Many previous approaches use long short-term memory networks with attention mechanisms to directly learn aspectspecific representations and model comment text. However, these methods always ignore the importance of the aspect terms position and interactive information between the aspect terms and other words. To address these issues, we propose an improved model based on convolutional neural networks. First, a novel relative position encode layer can integrate the relative position information of specific aspect terms validly in a text. Second, by using the aspect attention mechanism, the semantic relationship between aspect terms and words in the text is fully considered. To verify the effectiveness of the proposed models, we conduct a large number of experiments and comparisons on seven public datasets. The experimental results show that this model outperforms to other state-of-the-art methods. Keywords Aspect-based · CNN · Relative position · Aspect attention

B

Qingyu Xiong [email protected] Chao Wu [email protected] Min Gao [email protected] Qiude Li [email protected] Yang Yu [email protected] Kaige Wang [email protected]

1

Key Laboratory of Dependable Service Computing in Cyber Physical Society, Ministry of Education, Chongqing University, Chongqing, China

2

School of Big Data and Software Engineering, Chongqing University, Chongqing 400044, China

123

C. Wu et al.

1 Introduction Aspect-based sentiment analysis (ABSA) is a fine-grained task in the sentiment analysis. It aims to identify the sentiment tendency (e.g., positive, negative, neutral) of one specific aspect term or target entities in the text from human language [19,20,24]. Notably, in e-commerce platforms, a large number of users comment on different aspect terms of goods or services received, such as quality, price, logistics, and so on. For example, given a comment text “I bought a computer, its system is wonderful but the keyboard feels very bad,” there are two aspect opinion targets: “system” and “keyboard.” The reviewer has a positive sentiment on the “system” and a negative sentiment on the “keyboard.” General sentiment analysis (document level) methods cannot capture sentiment expressed in text very profoundly. Jiang et al. [10] manually evaluated multiple sentiment classifiers in the twitter text, and the results showed that 40% of the sentiment classification errors were caused by the lack of consideration of aspect information. Hence, ABSA for comment text is not only practical highly but