- Abstract: Generating coherent and cohesive long-form texts is a challenging problem in natural language generation. Previous works relied on a large amount of human-generated texts to train neural language models, however, few attempted to explicitly model the desired linguistic properties of natural language text, such as coherence and cohesion using neural networks. In this work, we train two expert discriminators for coherence and cohesion to provide hierarchical feedback for text generation. We also propose a simple variant of policy gradient, called 'negative-critical sequence training' in which the reward 'baseline' is constructed from randomly generated negative samples. We demonstrate the effectiveness of our approach through empirical studies, showing improvements over the strong baseline -- attention-based bidirectional MLE-trained neural language model -- in a number of automated metrics. The proposed model can serve as baseline architectures to promote further research in modeling additional linguistic properties for downstream NLP tasks.
- Keywords: text generation, natural language processing, neural language model
- TL;DR: We encode linguistic properties, such as, coherence and cohesion, into expert discriminators and improve text generation.