Comprehensive Analysis of Aspect Term Extraction Methods using Various Text Embeddings

TL;DR

This paper compares various aspect term extraction methods using different text embeddings and architectures, highlighting the impact of embedding coverage, source, and CRF layers on performance in aspect-based sentiment analysis.

Contribution

It provides a comprehensive comparison and ablation analysis of LSTM-based models with various embeddings and character extensions for aspect term extraction.

Findings

BiLSTM outperforms LSTM in aspect detection

Embedding coverage and source significantly affect performance

CRF layers consistently improve extraction results

Abstract

Recently, a variety of model designs and methods have blossomed in the context of the sentiment analysis domain. However, there is still a lack of wide and comprehensive studies of aspect-based sentiment analysis (ABSA). We want to fill this gap and propose a comparison with ablation analysis of aspect term extraction using various text embedding methods. We particularly focused on architectures based on long short-term memory (LSTM) with optional conditional random field (CRF) enhancement using different pre-trained word embeddings. Moreover, we analyzed the influence on the performance of extending the word vectorization step with character embedding. The experimental results on SemEval datasets revealed that not only does bi-directional long short-term memory (BiLSTM) outperform regular LSTM, but also word embedding coverage and its source highly affect aspect detection performance. An additional CRF layer consistently improves the results as well.