Forward and Backward Knowledge Transfer for Sentiment Classification

摘要: This paper studies the problem of learning a sequence sentiment classification tasks. The learned knowledge from each task is retained and used to help future or subsequent learning. paradigm called Lifelong Learning (LL). However, existing LL methods either only transfer forward do not go back improve model previous require training data retrain its exploit backward/reverse transfer. reverse in context naive Bayesian (NB) classification. It aims by leveraging without retraining using data. done exploiting key characteristic generative NB. That is, it possible NB classifier for improving parameters directly other Experimental results show that proposed method markedly outperforms baselines.