Language Model-Guided Knowledge Subgraphs for Question Answering

Abstract

Knowledge graphs for question answering can provide subgraphs based on different combinations of questions and answers for multiple reasoning chains, in which humans often find the answer for a question. In this paper, we introduce extracting multiple subgraphs fromKGs to model the reasoning process. We propose a new model to leverage language model-guided knowledge subgraphs, which explicitly provide potential multiple reasoning chains from different perspectives and are encoded with language models for joint reasoning. We evaluate our model in two datasets: Common-senseQA and OpenBookQA. The results show that the proposed approach outperforms state-of-the-art methods.

Tasks

Reproductions