MarkushGrapher: Joint Visual and Textual Recognition of Markush Structures
Lucas Morin, Valéry Weber, Ahmed Nassar, Gerhard Ingmar Meijer, Luc Van Gool, Yawei Li, Peter Staar · 2025
Read MarkushGrapher: Joint Visual and Textual Recognition of Markush Structures by Lucas Morin, Valéry Weber, Ahmed Nassar, Gerhard Ingmar Meijer, Luc Van Gool, Yawei Li, Peter Staar (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-16096.txt.
Description
Lucas Morin and co-authors present MarkushGrapher in 2025, a system that reads chemical template structures from patent documents. Markush structures show a core chemical form with variable groups, claiming whole families of compounds at once, and they matter for prior art search in drug and materials work. Their mixed text and image form has left them largely untouched by automatic extraction.
The method encodes text, image, and page layout together, plus a vision encoder tuned for chemical structures. It then writes out a graph form of the structure and a table of its variable groups. Lacking real training data, the team built a synthetic data pipeline that makes many realistic cases. They also release the first annotated benchmark of real Markush structures. Tests beat both chemistry specific and general vision language models in most settings.
Catalog record
| CREATOR | Lucas Morin, Valéry Weber, Ahmed Nassar, Gerhard Ingmar Meijer, Luc Van Gool, Yawei Li, Peter Staar |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-16096