Aspect-based Sentiment Classification with Aspect-specific Graph Convolutional Networks (EMNLP 2019)

- 8 mins

Abstract

1. Introduction

2. Background: Graph Convolutional Network

figure1

eq1

3. Aspect-specific Graph Convolutional Network

figure2

3.1 Embedding and Bidirectional LSTM

3.2 Obtaining Aspect-oriented Features

3.2.1 Graph Convolution over Dependency Trees

eq2

Position-aware Transformation (이해 잘 안됨)

eq456

3.2.2 Aspect-specific Masking

3.3 Aspect-aware Attention

eq89

3.4 Sentiment Classification

3.5 Training

4. Experiments

4.1 Datasets and Experimental Settings

table1

4.2 Models for Comparison

4.3 Results

table2

4.4 Ablation Study

table3

4.5 Case Study

table4

5. Discussion

figure34

5.1 Investigation on the Impact of GCN Layers

5.2 Investigation on the Effect of Multiple Aspects

7. Conclusions and Future Work

Joohong Lee

Joohong Lee

Machine Learning Researcher

rss facebook twitter github youtube mail spotify instagram linkedin google pinterest medium