Inquisitive Question Generation for High Level Text Comprehension

TL;DR

This paper introduces INQUISITIVE, a new dataset of high-level questions generated during reading, and evaluates GPT-2 based models for generating such questions, emphasizing the importance of context.

Contribution

The paper presents INQUISITIVE, a novel dataset of 19,000 high-level comprehension questions, and assesses the performance of GPT-2 models in generating contextually relevant inquisitive questions.

Findings

Readers use pragmatic strategies to seek information.

GPT-2 models can generate reasonable questions.

Context is crucial for high-quality question generation.

Abstract

Inquisitive probing questions come naturally to humans in a variety of settings, but is a challenging task for automatic systems. One natural type of question to ask tries to fill a gap in knowledge during text comprehension, like reading a news article: we might ask about background information, deeper reasons behind things occurring, or more. Despite recent progress with data-driven approaches, generating such questions is beyond the range of models trained on existing datasets. We introduce INQUISITIVE, a dataset of ~19K questions that are elicited while a person is reading through a document. Compared to existing datasets, INQUISITIVE questions target more towards high-level (semantic and discourse) comprehension of text. We show that readers engage in a series of pragmatic strategies to seek information. Finally, we evaluate question generation models based on GPT-2 and show that our model is able to generate reasonable questions although the task is challenging, and highlight the importance of context to generate INQUISITIVE questions.