The benefits of query-based KGQA systems for complex and temporal questions in LLM era

Аннотация

Large language models (LLMs) excel in question-answering(QA) but struggle with multi-hop reasoning and temporal questions.Query-based knowledge graph QA (KGQA) offers a modular alternativeby generating executable queries instead of direct answers.We explore anend-to-end query-based framework for WikiData QA, proposing a GPTbasedapproach that enhances performance on challenging multi-hop andtemporal benchmarks. Through generalization and rejection studies, weevaluate robustness across multi-hop and temporal QA datasets. Additionally,we introduce a novel entity linking and predicate matchingmethod using CoT reasoning. Our results demonstrate the potential ofquery-based KGQA pipelines for improving multi-hop and temporal QAwith small language models.Code: https://github.com/ar2max/NLDB-KGQA-System