Graph-Based Semi-Supervised Deep Learning for Indonesian Aspect-Based Sentiment Analysis

Product reviews on the marketplace are interesting to research. Aspect-based sentiment analysis (ABSA) can be used find in-depth information from a review. In one review, there several aspects with polarity of sentiment. Previous research has developed ABSA, but it still limitations in detecting and classification requires labeled data, obtaining data is very difficult. This graph-based semi-supervised approach improve ABSA. GCN GRN methods detect aspect opinion relationships. CNN RNN classification. A model was overcome data. The dataset an Indonesian-language review taken marketplace. small part manually, most automatically. experiment results for by comparing obtained best using method F1 score = 0.97144. 0.94020. Our label unlabeled automatically outperforms existing advanced models.