A Joint Model for Question Answering and Question Generation

TL;DR

This paper introduces a generative joint model for question answering and question generation that improves performance on the SQuAD dataset by learning both tasks simultaneously, advancing machine comprehension capabilities.

Contribution

The paper presents a novel sequence-to-sequence joint model that learns to ask and answer questions together, demonstrating improved empirical results over separate models.

Findings

Significant performance improvement on SQuAD dataset

Joint learning benefits both question answering and generation tasks

Proposes a new perspective on machine comprehension

Abstract

We propose a generative machine comprehension model that learns jointly to ask and answer questions based on documents. The proposed model uses a sequence-to-sequence framework that encodes the document and generates a question (answer) given an answer (question). Significant improvement in model performance is observed empirically on the SQuAD corpus, confirming our hypothesis that the model benefits from jointly learning to perform both tasks. We believe the joint model's novelty offers a new perspective on machine comprehension beyond architectural engineering, and serves as a first step towards autonomous information seeking.