Aspect Detection using Word and Char Embeddings with (Bi)LSTM and CRF

TL;DR

This paper introduces an aspect extraction method combining word and character embeddings with (Bi)LSTM and CRF, achieving state-of-the-art results on restaurant and laptop datasets.

Contribution

It demonstrates that combining word and character embeddings with (Bi)LSTM and CRF improves aspect detection accuracy over previous methods.

Findings

BiLSTM outperforms LSTM in aspect detection.

Adding CRF layer enhances model performance.

Word embedding coverage significantly affects results.

Abstract

We proposed a~new accurate aspect extraction method that makes use of both word and character-based embeddings. We have conducted experiments of various models of aspect extraction using LSTM and BiLSTM including CRF enhancement on five different pre-trained word embeddings extended with character embeddings. The results revealed that BiLSTM outperforms regular LSTM, but also word embedding coverage in train and test sets profoundly impacted aspect detection performance. Moreover, the additional CRF layer consistently improves the results across different models and text embeddings. Summing up, we obtained state-of-the-art F-score results for SemEval Restaurants (85%) and Laptops (80%).