Improving Adversarial Text Generation by Modeling the Distant Future
This addresses text generation quality for NLP applications, but appears incremental as it builds on existing imitation-learning and planning methods.
The paper tackles the problem of inconsistent semantics and limited fluency in long auto-regressive text generation by proposing a model-based imitation-learning approach with a guider network for longer-horizon planning, resulting in improved performance as demonstrated in experiments.
Auto-regressive text generation models usually focus on local fluency, and may cause inconsistent semantic meaning in long text generation. Further, automatically generating words with similar semantics is challenging, and hand-crafted linguistic rules are difficult to apply. We consider a text planning scheme and present a model-based imitation-learning approach to alleviate the aforementioned issues. Specifically, we propose a novel guider network to focus on the generative process over a longer horizon, which can assist next-word prediction and provide intermediate rewards for generator optimization. Extensive experiments demonstrate that the proposed method leads to improved performance.