Improving embedded knowledge graph multi-hop question answering by introducing relational chain reasoning

Knowledge Graph Question Answering (KGQA) aims to answer user-questions from a knowledge graph (KG) by identifying the reasoning relations between topic entity and answer. As complex branch task of KGQA, multi-hop KGQA requires over relational chain preserved in KG arrive at right Despite recent successes, existing works on answering questions still face following challenges: (i) The absence an explicit order reflected user-question stems misunderstanding user’s intentions. (ii) Incorrectly capturing types weak supervision which dataset lacks intermediate annotations due expensive labeling cost. (iii) Failing consider implicit implied structured because limited neighborhoods size constraint subgraph retrieval-based algorithms. To address these issues we propose novel model herein, namely Relational Chain based Embedded (Rce-KGQA), simultaneously utilizes revealed natural language question stored KG. Our extensive empirical study three open-domain benchmarks proves that our method significantly outperforms state-of-the-art counterparts like GraftNet, PullNet EmbedKGQA. Comprehensive ablation experiments also verify effectiveness task. We have made model’s source code available github: https://github.com/albert-jin/Rce-KGQA .