Semi-supervised Text Regression with Conditional Generative Adversarial Networks

Abstract

Enormous online textual information provides intriguing opportunities for\nunderstandings of social and economic semantics. In this paper, we propose a\nnovel text regression model based on a conditional generative adversarial\nnetwork (GAN), with an attempt to associate textual data and social outcomes in\na semi-supervised manner. Besides promising potential of predicting\ncapabilities, our superiorities are twofold: (i) the model works with\nunbalanced datasets of limited labelled data, which align with real-world\nscenarios; and (ii) predictions are obtained by an end-to-end framework,\nwithout explicitly selecting high-level representations. Finally we point out\nrelated datasets for experiments and future research directions.\n