Uma Ferramenta Web Baseada em LLMs para Respostas a Perguntas Simples sobre Bases de Conhecimento Heterogêneas
Resumo
We present KIF-QA, a Web tool for answering simple questions over heterogeneous knowledge graphs. The Web tool is built on top of a homonymous framework, KIF-QA, that uses in-context learning over off-the-shelf pre-trained large language models for semantic parsing. Because it relies on in-context (few-shot) learning to generate logical forms and disambiguate entities, and because it uses KIF (a knowledge integration framework based on Wikidata) to mediate all access to the target knowledge graphs, KIF-QA’s approach is general. It requires neither training nor fine-tuning and can be easily adapted to work with different graphs. The KIF-QA Web tool allows the user to change the target graphs (Wikidata, DBpedia, or PubChem) and the language model used. The user can also view and modify the details of each step taken by KIF-QA to generate the answers to the question. Both the Web tool and the KIF-QA framework are released as open-source.
Referências
J. Berant, A. Chou, R. Frostig, and P. Liang. 2013. Semantic Parsing on Freebase from Question-Answer Pairs. In Proc. 2013 Conf. Empirical Methods in Natural Language Processing. ACL, 1533–1544.
X. Huang, S. Cheng, Y. Shu, Y. Bao, and Y. Qu. 2023. Question Decomposition Tree for Answering Complex Questions over Knowledge Bases. In Proc. AAAI Conf. Artificial Intelligence, Vol. 37. 12924–12932.
S. Kim, J. Chen, T. Cheng, A. Gindulyte, J. He, S. He, Q. Li, B .A. Shoemaker, P. A. Thiessen, B. Yu, L. Zaslavsky, J. Zhang, and E. E. Bolton. 2023. PubChem 2023 Update. Nucleic Acids Res. 51, D1 (October 2023), D1373–D1380.
Y. Lan, G. He, J. Jiang, J. Jiang, W. X. Zhao, and J.-R. Wen. 2023. Complex Knowledge Base Question Answering: A Survey. IEEE Trans. Know. Data Eng. 35, 11 (2023), 11196–11215.
J. Lehmann, R. Isele, M. Jakob, A. Jentzsch, D. Kontokostas, P. N. Mendes, S. Hellmann, M. Morsey, P. van Kleef, S. Auer, and C. Bizer. 2015. DBpedia - A largescale, multilingual knowledge base extracted from Wikipedia. Semant. Web 6 (2015), 167–195.
G. Lima, J. M. B. Rodrigues, M. Machado, E. Soares, S. R. Fiorini, R. Thiago, L. G. Azevedo, V. T. da Silva, and R. Cerqueira. 2024. KIF: A Wikidata-Based Framework for Integrating Heterogeneous Knowledge Sources. arXiv:2403.10304
M. Machado, J. P. P. Campos, G. Lima, and V. T. da Silva. 2025. KIF-QA: Using Off-the-shelf LLMs to Answer Simple Questions over Heterogeneous Knowledge Bases. In Joint Proc. 5th Wikidata Workshop for the Scientific Wikidata Community co-located with the 24th International Semantic Web Conference (ISWC 2025), Nara, Japan, November 2–6, 2025.
A. Razzhigaev, M. Salnikov, V. Malykh, P. Braslavski, and A. Panchenko. 2023. A System for Answering Simple Questions in Multiple Languages. In Proc. 61st Annual Meeting of ACL (Volume 3: System Demonstrations). ACL, 524–537.
D. Vrandečić and M. Krötzsch. 2014. Wikidata: A Free Collaborative Knowledgebase. Commun. ACM 57, 10 (September 2014), 78–85.
W3C RDF Working Group. 2014. Resource Description Framework (RDF): Concepts and Abstract Syntax. W3C Recommendation. W3C.
W3C SPARQL Working Group. 2013. SPARQL 1.1 Overview. W3C Recommendation. W3C.
