Multi-label Text Classification and Text Adversarial Attack

Abstract

Multi-label classification is an extension of multi-class classification. For multi-label problem, each instance may not be restricted to have only one label. In this paper, the methods to solve multi-label classification are divided into four aspects which are binary relevance method, label combination method, classifier chain and ensemble classifier chain. In order to enhance the performance of the text classifier, text adversarial attack should be used to enrich the training dataset. Thus, the related works with text adversarial attack are also introduced. In the end, we also explore some potential future issues in multi-label text classification and text adversarial attack.