Generating Factoid Questions With Recurrent Neural Networks: The 30M Factoid Question-Answer Corpus

TL;DR

This paper introduces a large-scale 30 million factoid question-answer corpus generated by a neural network from Freebase, enabling advancements in machine learning for question answering.

Contribution

It presents a novel neural network architecture for converting knowledge base facts into natural language questions, creating the largest question-answer corpus to date.

Findings

The neural network outperforms template-based baselines in automatic evaluations.

Generated questions are comparable in quality to human questions according to human evaluators.

The corpus facilitates large-scale supervised learning for question-answering systems.

Abstract

Over the past decade, large-scale supervised learning corpora have enabled machine learning researchers to make substantial advances. However, to this date, there are no large-scale question-answer corpora available. In this paper we present the 30M Factoid Question-Answer Corpus, an enormous question answer pair corpus produced by applying a novel neural network architecture on the knowledge base Freebase to transduce facts into natural language questions. The produced question answer pairs are evaluated both by human evaluators and using automatic evaluation metrics, including well-established machine translation and sentence similarity metrics. Across all evaluation criteria the question-generation model outperforms the competing template-based baseline. Furthermore, when presented to human evaluators, the generated questions appear comparable in quality to real human-generated questions.