Multi-task Solution for Aspect Category Sentiment Analysis on Vietnamese Datasets

Abstract

In this article, we solved two tasks in the Vietnamese Aspect-based Sentiment Analysis problem: Aspect Category Detection (ACD) and Sentiment Polarity Classification (SPC). Besides, we proposed end-to-end models to handle the above tasks simultaneously for two domains (Restaurant and Hotel) in the VLSP 2018 Aspect-based Sentiment Analysis dataset using PhoBERT as Pre-trained language models for Vietnamese in two ways: Multi-task and Multi-task with Multi-branch approach. Both models give very good results when applied preprocessing. Specifically, the Multi-task model achieves state-of-the-art (SOTA) results in the Hotel domain of the VLSP 2018 ABSA dataset, with the F1-score being 82.55% for ACD and 77.32% for ACD with SPC. For the Restaurant domain, our Multi-task model also achieved SOTA in the ACD with SPC task by an F1-score of 71.55% and 83.29% for the ACD.