Learning Interpretable and Discrete Representations with Adversarial Training for Unsupervised Text Classification
This addresses the challenge of interpretable text analysis for researchers and practitioners by enabling unsupervised classification with coherent topics, though it is incremental as it builds on existing adversarial and representation learning techniques.
The paper tackles the problem of learning discrete and interpretable representations from unlabeled text for unsupervised classification, proposing TIGAN which encodes texts into disentangled discrete and continuous parts, achieving superior performance on six corpora compared to unsupervised baselines and matching a weakly-supervised method.
Learning continuous representations from unlabeled textual data has been increasingly studied for benefiting semi-supervised learning. Although it is relatively easier to interpret discrete representations, due to the difficulty of training, learning discrete representations for unlabeled textual data has not been widely explored. This work proposes TIGAN that learns to encode texts into two disentangled representations, including a discrete code and a continuous noise, where the discrete code represents interpretable topics, and the noise controls the variance within the topics. The discrete code learned by TIGAN can be used for unsupervised text classification. Compared to other unsupervised baselines, the proposed TIGAN achieves superior performance on six different corpora. Also, the performance is on par with a recently proposed weakly-supervised text classification method. The extracted topical words for representing latent topics show that TIGAN learns coherent and highly interpretable topics.