Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Joint Learning of Local and Global Features for Aspect-based Sentiment Classification

Hao Niu, Yun Xiong, Xiaosu Wang, Philip S. Yu

TL;DR

This work tackles aspect-based sentiment classification by jointly modeling local and global information. It introduces Gaussian mask and covariance self-attention to adaptively capture local context around the given aspect term, and AWIG plus DGAT to exploit long-distance dependency-relations-driven global information. Empirical results on SemEval 2014 and Twitter show state-of-the-art performance and robust ablations confirm the contribution of each component. The approach highlights the value of incorporating dependency relation tags and edge semantics for effective ASC in real-world texts.

Abstract

Aspect-based sentiment classification (ASC) aims to judge the sentiment polarity conveyed by the given aspect term in a sentence. The sentiment polarity is not only determined by the local context but also related to the words far away from the given aspect term. Most recent efforts related to the attention-based models can not sufficiently distinguish which words they should pay more attention to in some cases. Meanwhile, graph-based models are coming into ASC to encode syntactic dependency tree information. But these models do not fully leverage syntactic dependency trees as they neglect to incorporate dependency relation tag information into representation learning effectively. In this paper, we address these problems by effectively modeling the local and global features. Firstly, we design a local encoder containing: a Gaussian mask layer and a covariance self-attention layer. The Gaussian mask layer tends to adjust the receptive field around aspect terms adaptively to deemphasize the effects of unrelated words and pay more attention to local information. The covariance self-attention layer can distinguish the attention weights of different words more obviously. Furthermore, we propose a dual-level graph attention network as a global encoder by fully employing dependency tag information to capture long-distance information effectively. Our model achieves state-of-the-art performance on both SemEval 2014 and Twitter datasets.

Joint Learning of Local and Global Features for Aspect-based Sentiment Classification

TL;DR

This work tackles aspect-based sentiment classification by jointly modeling local and global information. It introduces Gaussian mask and covariance self-attention to adaptively capture local context around the given aspect term, and AWIG plus DGAT to exploit long-distance dependency-relations-driven global information. Empirical results on SemEval 2014 and Twitter show state-of-the-art performance and robust ablations confirm the contribution of each component. The approach highlights the value of incorporating dependency relation tags and edge semantics for effective ASC in real-world texts.

Abstract

Aspect-based sentiment classification (ASC) aims to judge the sentiment polarity conveyed by the given aspect term in a sentence. The sentiment polarity is not only determined by the local context but also related to the words far away from the given aspect term. Most recent efforts related to the attention-based models can not sufficiently distinguish which words they should pay more attention to in some cases. Meanwhile, graph-based models are coming into ASC to encode syntactic dependency tree information. But these models do not fully leverage syntactic dependency trees as they neglect to incorporate dependency relation tag information into representation learning effectively. In this paper, we address these problems by effectively modeling the local and global features. Firstly, we design a local encoder containing: a Gaussian mask layer and a covariance self-attention layer. The Gaussian mask layer tends to adjust the receptive field around aspect terms adaptively to deemphasize the effects of unrelated words and pay more attention to local information. The covariance self-attention layer can distinguish the attention weights of different words more obviously. Furthermore, we propose a dual-level graph attention network as a global encoder by fully employing dependency tag information to capture long-distance information effectively. Our model achieves state-of-the-art performance on both SemEval 2014 and Twitter datasets.
Paper Structure (32 sections, 24 equations, 7 figures, 4 tables)

This paper contains 32 sections, 24 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Attention-based Models
  4. Graph-based Models
  5. Proposed Method
  6. Input Representation Layer
  7. Local Encoder
  8. Gaussian Mask Layer
  9. Covariance Self-attention
  10. Proposition:
  11. Proof Sketch :
  12. Detailed Proof of Proposition:
  13. Global Encoder
  14. Aspect-word Interactive Graph (AWIG)
  15. Dual-level graph attention network (DGAT)
  16. ...and 17 more sections

Figures (7)

  • Figure 1: Overall architecture of proposed GDD.
  • Figure 2: Construction of an aspect-word interactive graph (AWIG) (bottom) from an original dependency parse tree (top). The aspect term is noise level with the sentiment polarity negative.
  • Figure 3: Overview of the proposed DGAT (a). DGAT comprises two sorts of multi-heads attention mechanism: dual-level attention head (b) and relational head.
  • Figure 4: Gaussian mask visualization of typical cases.
  • Figure 5: Analysis of local context receptive field on the Restaurant and Laptop datasets.
  • ...and 2 more figures