A Large-Scale Dataset for Molecular Structure-Language Description via a Rule-Regularized Method

TL;DR

The paper tackles the problem of aligning molecular structure with natural language for robust chemical reasoning by introducing a fully automated annotation pipeline that extends OPSIN to produce enriched structural metadata. This metadata grounds LLM-generated descriptions, enabling large-scale creation of structure-description pairs (~$163{,}085$) with a validation precision of $98.6\%$ on a subset of $2{,}000$ samples. Key contributions include metadata augmentation to capture fused/bridged/spiro topology, a metadata-guided prompting strategy, and a hybrid validation scheme that combines LLM and expert checks. The resulting dataset provides a scalable, reliable foundation for molecule-language alignment and is readily extensible to larger datasets and broader chemical tasks that rely on precise structural descriptions.

Abstract

Molecular function is largely determined by structure. Accurately aligning molecular structure with natural language is therefore essential for enabling large language models (LLMs) to reason about downstream chemical tasks. However, the substantial cost of human annotation makes it infeasible to construct large-scale, high-quality datasets of structure-grounded descriptions. In this work, we propose a fully automated annotation framework for generating precise molecular structure descriptions at scale. Our approach builds upon and extends a rule-based chemical nomenclature parser to interpret IUPAC names and construct enriched, structured XML metadata that explicitly encodes molecular structure. This metadata is then used to guide LLMs in producing accurate natural-language descriptions. Using this framework, we curate a large-scale dataset of approximately $163$k molecule-description pairs. A rigorous validation protocol combining LLM-based and expert human evaluation on a subset of $2,000$ molecules demonstrates a high description precision of $98.6\%$. The resulting dataset provides a reliable foundation for future molecule-language alignment, and the proposed annotation method is readily extensible to larger datasets and broader chemical tasks that rely on structural descriptions.

Figures (6)

• Figure 1: An illustrative example motivating this work. Existing approaches align molecular representations with high-level objectives, while we argue molecule-language alignment should be structure-grounded, with higher-level reasoning handled by the LLM backbone, analogous to image-language alignment. Real molecular descriptions in this work are substantially more complex than this example.
• Figure 2: Illustrative example of the molecule (7'R)-7'-methyl-7-((E)-prop-1-en-1-yl)-5',6'-dihydrospiro[benzo[e][1,2]oxazine-4,4'-[2,5]methanocyclopenta[b]furan]. The top shows the decomposition from basic components to the complete structure. The bottom presents the structure metadata constructed by our approach; the corresponding native OPSIN XML output is shown in Appendix Fig.\ref{['fig:running_example_opsin_xml']} for comparison. The natural-language structural description generated from this metadata is provided in Appendix Fig.\ref{['fig:running_example_descriptions']}.
• Figure S1: Molecular structure and generated natural-language structural description for the illustrative example shown in Fig.\ref{['fig:main_example']}. The structured metadata provided as input to the LLM are illustrated in Fig.\ref{['fig:main_example']}.
• Figure S2: Molecular structure XML parse tree produced by the native OPSIN tool opsin after the structure assembly for the illustrative example shown in Fig.\ref{['fig:main_example']}. The corresponding XML representation generated by our approach is shown in Fig.\ref{['fig:main_example']} for comparison.
• Figure S3: Distribution statistics of the entire generated dataset, including non-hydrogen atom counts and description word counts, across easy, medium, and hard categories.