THANOS: The Aspect Classification Model for Imbalanced Vietnamese E-commerce Review Data

This article introduces several methods for aspect classification using machine learning models for data collected from customers' reviews on two e-commerce sites and mainly focuses on handling data imbalance to improve classifier performance. To this end, we describe the problem as a binary classifier at the sentence level. Sentences will be expressed as feature vectors using One-hot combined with Chi-square statistics, and we will use basic machine learning models such as Naive Bayes, SVM, Random Forest, Linear Regression for training before classifying the aspects included in. We tested the strategy on a dataset generated in Vietnamese based on the comments from users who purchased products on e-commerce sites. The experimental results suggest that the method produces better outcomes than before the imbalance handling.