The article introduces DECIMER, an open-source artificial intelligence platform that automates the conversion of chemical structural formulas from images into machine-readable codes. This innovation addresses a critical bottleneck in cheminformatics, where the intuitive visual language of chemistry remains inaccessible to automated databases because computers only perceive visual data as pixel arrays rather than semantic chemical information. By leveraging deep learning models trained on hundreds of millions of existing structures, DECIMER can instantly identify and translate structural formulas embedded in scientific articles and patents. This capability allows researchers and companies to efficiently populate searchable databases with previously locked-away visual data, significantly accelerating the integration of historical and current scientific literature into digital formats for further analysis and discovery. This development is vital for open data initiatives because it unlocks vast amounts of legacy chemical knowledge that was previously trapped in non-searchable image formats. By making this information computable and freely accessible, DECIMER supports the sustainable preservation of scientific heritage and enhances global collaboration, ensuring that vital chemical data is not lost but is instead available for reuse and integration into open research infrastructures.
Source: sciencedaily.comPublished on 2023-08-23
Related news
- NVIDIA DLSS 3.5 uses a mountain of data and AI to improve ray tracing on RTX GPUs
- EE.UU. critica a China sobre la transparencia de sus informes de datos económicos
- Drug deaths in Scotland down for second year in a row - but rich-poor gap widens
- Rechazan petición de proteger con derechos de autor obra de arte hecha con IA en EE. UU. - Pulzo