Supervised Gradual Machine Learning for Aspect-Term Sentiment Analysis

Abstract Recent work has shown that Aspect-Term Sentiment Analysis (ATSA) can be effectively performed by Gradual Machine Learning (GML). However, the performance of current unsupervised solution is limited inaccurate and insufficient knowledge conveyance. In this paper, we propose a supervised GML approach for ATSA, which exploit labeled training data to improve It leverages binary polarity relations between instances, either similar or opposite, enable Besides explicit indicated discourse structures, it also separately supervises classification DNN Siamese network extract implicit relations. The proposed fulfills conveyance modeling detected as features in factor graph. Our extensive experiments on real benchmark show achieves state-of-the-art across all test workloads. demonstrates clearly that, collaboration with feature extraction, outperforms pure solutions.