Aspect Term Extraction Using Deep Learning-Based Approach on Indonesian Restaurant Reviews

Aspect term extraction is a fundamental process in aspect-based sentiment analysis. Aspect term extraction aims to identify the review text span that contains the aspect mentions. In this paper, we present our work on aspect term extraction for Indonesian restaurant reviews, using a deep learning-based approach. We collected and annotated an Indonesian restaurant reviews dataset, obtained from a restaurant review website. We performed the annotation at a token-level and used the following aspect labels to annotate the reviews: FOOD, PRICE, AMBIENCE, SERVICE, and MISCELLANEOUS. This paper treats aspect extraction as a token-level classification. We employed a Convolutional Neural Network (CNN) model and Long Short-Term Memory (LSTM) model for the classification. The experimental result showed that the LSTM method gives the best performance, with the micro average F1-score is 55,1%.