BigDipperAI at LLMs4OL 2026 Task Flagship: Structured Generation of Primitive Ontology Triples with Schema-Aligned Auxiliary Supervision
Open Conference Proceedings · 03.11.2026 · 11
DOI: 10.52825/ocp.v11i.3657 ↗Kurz gefasst
Fachartikel in der Zeitschrift Open Conference Proceedings, online erschienen am 03.11.2026. Verfasst von Guojian Zhao und 2 weiteren Autor:innen. Die Arbeit nennt 16 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Primitive ontology learning transforms natural-language contexts into typed subject–relation–object structures for machine-readable knowledge organization and reuse. The LLMs4OL 2026 Task Flagship requires integrated extraction of terms, types, taxonomies, and relations from raw text. Our submission converts Task Reuse records into schema-aligned auxiliary supervision for target-task training. Our approach adapts Qwen3-8B with relation-aware weighted four-bit QLoRA to generate primitive ontology triples directly from text, followed by deterministic recovery and validation. Organizer-issued evaluation yielded exact, fuzzy, and semantic Graph Similarity scores of 0.6487, 0.7932, and 0.8258, averaging 0.7559. The evaluation shows higher agreement under semantic matching, while non-taxonomic relation extraction remains the clearest weakness.
Zitieren
APA
Zhao, G., Tang, Y., & Li, Z. (2026). BigDipperAI at LLMs4OL 2026 Task Flagship: Structured Generation of Primitive Ontology Triples with Schema-Aligned Auxiliary Supervision. Open Conference Proceedings, 11. https://doi.org/10.52825/ocp.v11i.3657
Vancouver
Zhao G, Tang Y, Li Z. BigDipperAI at LLMs4OL 2026 Task Flagship: Structured Generation of Primitive Ontology Triples with Schema-Aligned Auxiliary Supervision. Open Conference Proceedings. 2026;11. doi:10.52825/ocp.v11i.3657
BibTeX
@article{Zhao2026,
author = {Zhao, Guojian and Tang, Yuxuan and Li, Zhao},
title = {BigDipperAI at LLMs4OL 2026 Task Flagship: Structured Generation of Primitive Ontology Triples with Schema-Aligned Auxiliary Supervision},
journal = {Open Conference Proceedings},
year = {2026},
volume = {11},
doi = {10.52825/ocp.v11i.3657}
}
EUVIMED