CLASSIC: Continual and Contrastive Learning of Aspect Sentiment Classification Tasks

This paper studies continual learning (CL) of a sequence of aspect sentiment\nclassification(ASC) tasks in a particular CL setting called domain incremental\nlearning (DIL). Each task is from a different domain or product. The DIL\nsetting is particularly suited to ASC because in testing the system needs not\nknow the task/domain to which the test data belongs. To our knowledge, this\nsetting has not been studied before for ASC. This paper proposes a novel model\ncalled CLASSIC. The key novelty is a contrastive continual learning method that\nenables both knowledge transfer across tasks and knowledge distillation from\nold tasks to the new task, which eliminates the need for task ids in testing.\nExperimental results show the high effectiveness of CLASSIC.\n