Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answering

TL;DR

This paper presents a graph-based recurrent retrieval method that learns to identify reasoning paths over Wikipedia to improve multi-hop question answering, achieving state-of-the-art results on multiple datasets.

Contribution

Introduces a novel recurrent neural network retriever that learns to sequentially retrieve evidence paths over Wikipedia for multi-hop QA.

Findings

Achieves state-of-the-art results on three open-domain QA datasets.

Outperforms previous models by over 14 points on HotpotQA.

Demonstrates robustness and effectiveness of the reasoning path retrieval approach.

Abstract

Answering questions that require multi-hop reasoning at web-scale necessitates retrieving multiple evidence documents, one of which often has little lexical or semantic relationship to the question. This paper introduces a new graph-based recurrent retrieval approach that learns to retrieve reasoning paths over the Wikipedia graph to answer multi-hop open-domain questions. Our retriever model trains a recurrent neural network that learns to sequentially retrieve evidence paragraphs in the reasoning path by conditioning on the previously retrieved documents. Our reader model ranks the reasoning paths and extracts the answer span included in the best reasoning path. Experimental results show state-of-the-art results in three open-domain QA datasets, showcasing the effectiveness and robustness of our method. Notably, our method achieves significant improvement in HotpotQA, outperforming the previous best model by more than 14 points.