Putting the Horse Before the Cart:A Generator-Evaluator Framework for Question Generation from Text
This work addresses the problem of generating high-quality questions from text for NLP applications, representing a strong specific gain rather than a foundational advancement.
The authors tackled automatic question generation from text by proposing a generator-evaluator framework that optimizes semantics and structure, resulting in significant outperformance over state-of-the-art systems on the SQuAD benchmark in both automatic and human evaluations.
Automatic question generation (QG) is a useful yet challenging task in NLP. Recent neural network-based approaches represent the state-of-the-art in this task. In this work, we attempt to strengthen them significantly by adopting a holistic and novel generator-evaluator framework that directly optimizes objectives that reward semantics and structure. The {\it generator} is a sequence-to-sequence model that incorporates the {\it structure} and {\it semantics} of the question being generated. The generator predicts an answer in the passage that the question can pivot on. Employing the copy and coverage mechanisms, it also acknowledges other contextually important (and possibly rare) keywords in the passage that the question needs to conform to, while not redundantly repeating words. The {\it evaluator} model evaluates and assigns a reward to each predicted question based on its conformity to the {\it structure} of ground-truth questions. We propose two novel QG-specific reward functions for text conformity and answer conformity of the generated question. The evaluator also employs structure-sensitive rewards based on evaluation measures such as BLEU, GLEU, and ROUGE-L, which are suitable for QG. In contrast, most of the previous works only optimize the cross-entropy loss, which can induce inconsistencies between training (objective) and testing (evaluation) measures. Our evaluation shows that our approach significantly outperforms state-of-the-art systems on the widely-used SQuAD benchmark as per both automatic and human evaluation.