Unsupervised Multi-hop Question Answering by Question Generation

TL;DR

This paper introduces MQA-QG, an unsupervised framework for multi-hop question answering that generates training data from various sources, enabling effective QA models without human-labeled multi-hop questions.

Contribution

The paper presents a novel unsupervised method for generating multi-hop QA training data, reducing reliance on human annotations and improving training efficiency.

Findings

Achieves 61% of supervised performance on HybridQA

Achieves 83% of supervised performance on HotpotQA

Pretraining with generated data reduces need for human annotations

Abstract

Obtaining training data for multi-hop question answering (QA) is time-consuming and resource-intensive. We explore the possibility to train a well-performed multi-hop QA model without referencing any human-labeled multi-hop question-answer pairs, i.e., unsupervised multi-hop QA. We propose MQA-QG, an unsupervised framework that can generate human-like multi-hop training data from both homogeneous and heterogeneous data sources. MQA-QG generates questions by first selecting/generating relevant information from each data source and then integrating the multiple information to form a multi-hop question. Using only generated training data, we can train a competent multi-hop QA which achieves 61% and 83% of the supervised learning performance for the HybridQA and the HotpotQA dataset, respectively. We also show that pretraining the QA system with the generated data would greatly reduce the demand for human-annotated training data. Our codes are publicly available at https://github.com/teacherpeterpan/Unsupervised-Multi-hop-QA.