Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Leveraging Deep Neural Networks for Aspect-Based Sentiment Classification

Chen Li, Debo Cheng, Yasuhiko Morimoto

TL;DR

The paper tackles aspect-based sentiment analysis by addressing the information loss that can occur when using standard GCNs for syntactic features. It introduces EEGCN, a model that fuses Bi-LSTM, a Transformer encoder, and an edge-enhanced Bi-GCN built on a dependency tree, complemented by aspect-specific masking to focus on relevant signals. Across four benchmark datasets, EEGCN consistently outperforms baselines and shows that edge-aware graph propagation and global context integration significantly improve ABSA accuracy and F1 scores. The work highlights the value of incorporating syntactic edges with meaningful weights and bidirectional message passing, while outlining avenues for future improvements in trainable edge representations and global-information fusion.

Abstract

Aspect-based sentiment analysis seeks to determine sentiment with a high level of detail. While graph convolutional networks (GCNs) are commonly used for extracting sentiment features, their straightforward use in syntactic feature extraction can lead to a loss of crucial information. This paper presents a novel edge-enhanced GCN, called EEGCN, which improves performance by preserving feature integrity as it processes syntactic graphs. We incorporate a bidirectional long short-term memory (Bi-LSTM) network alongside a self-attention-based transformer for effective text encoding, ensuring the retention of long-range dependencies. A bidirectional GCN (Bi-GCN) with message passing then captures the relationships between entities, while an aspect-specific masking technique removes extraneous information. Extensive evaluations and ablation studies on four benchmark datasets show that EEGCN significantly enhances aspect-based sentiment analysis, overcoming issues with syntactic feature extraction and advancing the field's methodologies.

Leveraging Deep Neural Networks for Aspect-Based Sentiment Classification

TL;DR

The paper tackles aspect-based sentiment analysis by addressing the information loss that can occur when using standard GCNs for syntactic features. It introduces EEGCN, a model that fuses Bi-LSTM, a Transformer encoder, and an edge-enhanced Bi-GCN built on a dependency tree, complemented by aspect-specific masking to focus on relevant signals. Across four benchmark datasets, EEGCN consistently outperforms baselines and shows that edge-aware graph propagation and global context integration significantly improve ABSA accuracy and F1 scores. The work highlights the value of incorporating syntactic edges with meaningful weights and bidirectional message passing, while outlining avenues for future improvements in trainable edge representations and global-information fusion.

Abstract

Aspect-based sentiment analysis seeks to determine sentiment with a high level of detail. While graph convolutional networks (GCNs) are commonly used for extracting sentiment features, their straightforward use in syntactic feature extraction can lead to a loss of crucial information. This paper presents a novel edge-enhanced GCN, called EEGCN, which improves performance by preserving feature integrity as it processes syntactic graphs. We incorporate a bidirectional long short-term memory (Bi-LSTM) network alongside a self-attention-based transformer for effective text encoding, ensuring the retention of long-range dependencies. A bidirectional GCN (Bi-GCN) with message passing then captures the relationships between entities, while an aspect-specific masking technique removes extraneous information. Extensive evaluations and ablation studies on four benchmark datasets show that EEGCN significantly enhances aspect-based sentiment analysis, overcoming issues with syntactic feature extraction and advancing the field's methodologies.

Paper Structure

This paper contains 17 sections, 18 equations, 4 figures, 3 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Aspect-based Sentiment Classification
  4. RNN-based Sentiment Classification
  5. GCN-based Sentiment Classification
  6. Model
  7. Architecture Overview
  8. EEGCN
  9. Training Phase
  10. Experiments
  11. Datasets
  12. Experimental Setup
  13. Metrics
  14. Results
  15. Effect of GCN Layers
  16. ...and 2 more sections

Figures (4)

  • Figure 1: An example of spurious associations of a customer review system.
  • Figure 2: The EEGCN architecture for aspect-based sentiment analysis consists of the following key components: (a) The word embedding layer converts sentences into vector representations, forming the basis for further processing. (b) The Bi-LSTM network captures contextual relationships within the sentence to provide a deeper understanding of word interactions. (c) The transformer layer, with its self-attention mechanism, analyzes global word features and interrelationships in lengthy texts, enhancing contextual comprehension. (d) The dependency parsing layer constructs a dependency tree to represent grammatical structures and word dependencies. (e) The Bi-GCN utilizes message passing to propagate information across the nodes of the dependency tree, effectively modeling word relationships. (f) The aspect-specific masking technique refines representations by isolating aspect-related information, reducing redundancy and improving analysis accuracy. (g) The classification layer categorizes the sentiment of the sentence into predefined categories, such as positive, negative, or neutral.
  • Figure 3: Change curve of Acc scores with the number of GCN layers.
  • Figure 4: Change curve of F1 scores with the number of GCN layers.