Chemistry buried in patent drawings, scientific figures and reaction schemes is being brought into searchable databases through a new AI collaboration between Elsevier and LG AI Research.

 

The companies said that chemistry-specific AI vision technology developed by LG AI Research is now being used in Elsevier’s content extraction and curation workflows for Reaxys, its chemistry discovery platform. The technology is designed to identify and extract substances and other chemical information that previously existed primarily as images rather than searchable text.

 

For researchers, the change could reduce the need to manually inspect documents when searching for compounds, structures and other chemical evidence.

 

“Every hour a chemist spends deciphering figures or images to see what has already been made is an hour that could instead be spent on chemistry discovery. Our partnership with LG AI Research gives that time back, lifting more chemistry out of the image and into Reaxys - curated, searchable and ready to act on. A structure buried in a figure should be evidence rather than a dead end," said Mirit Eldor, Managing Director, Life Sciences, Elsevier.

 

Chemical information is often communicated visually. Molecular structures, bonds, stereochemistry and reaction schemes can carry information that does not appear in a document’s searchable text.

 

That creates a major indexing challenge. A missed bond or incorrectly interpreted structure can lead to the wrong compound being identified, while structures that cannot be extracted at all may remain effectively invisible to conventional search.

 

Elsevier said the challenge is particularly relevant to areas including novelty searching, competitive intelligence and synthesis planning, as well as inorganic and organometallic chemistry, where structures can be especially complex.

 

LG AI Research’s system combines molecule detection, reaction-diagram parsing and optical chemical structure recognition (OCSR) in a single AI model. According to the company, its published benchmarking shows the model outperforming alternatives in extracting chemistry from complete document pages.

 

Before being deployed, extracted information is checked against existing Reaxys benchmarks, with the companies saying the pipeline underwent extensive testing across Elsevier’s data and workflows.

 

“Understanding scientific images requires AI engineered specifically for chemistry — where every bond and spatial layout holds critical meaning. We designed our AI vision model to decode these complex visual representations with human-expert precision. 

 

"Through our integration with Elsevier, we are proving that advanced chemistry-specific AI can seamlessly convert raw visual data into structured knowledge for global researchers," said Hwayoung Edward Lee, lead of the AI Biz Transformation Unit at LG AI Research.

 

The collaboration is now moving beyond individual chemical substances. Reaction extraction is the next stage, with the companies working to capture reaction information from images and expand the evidence available through Reaxys.

 

Elsevier and LG AI Research also said they are exploring additional customer challenges where LG AI Research’s AI capabilities could be combined with Elsevier’s chemistry content, scientific expertise and curation.