Semi-supervised Review-Aware Rating Regression (Student Abstract)

Semi-supervised learning is a promising solution to mitigate data sparsity in review-aware rating regression (RaRR), but it bears the risk of with noisy pseudo-labelled data. In this paper, we propose paradigm called co-training-teaching (CoT2), which integrates merits both co-training and co-teaching towards robust semi-supervised RaRR. Concretely, CoT2 employs two predictors each them alternately plays roles "labeler" "validator" generate validate instances. Extensive experiments show that considerably outperforms state-of-the-art RaRR techniques, especially when training severely insufficient.