Google created machine annotations for scientific content of the world-wide patent corpus, and released the data publicly [1, 2]. This dataset includes billions of mentions of chemical compounds, diseases, proteins, genes, anatomy, etc, extracted from the full text, images, PDFs, and translations of 130 million patents. We are actively working on extending data analytics utilized in PubChem to the patent data.<br/><br/>The literature knowledge panels developed and implemented in PubChem help to uncover and summarize important relationships between chemicals, genes, proteins, and diseases by analyzing co-occurrences of terms in biomedical literature abstracts [3]. We are working on extending this useful analysis and visualization technique to the patent dataset. While named entities in PubMed records are recognized using NLP software in [3], entities annotated by Google are used in this project. They are matched to chemical compounds in PubChem, disease names in Medical Subject Headings (MeSH), and gene/protein names in popular gene/protein information resources, and the most closely related entities are identified using statistical analysis and relevance-based sampling.<br/><br/>There is a significant variation in informativeness of annotated entities in Google Patents annotations due to the number of entities annotated in patent parts. This requires a fundamental rework of the relevance and co-occurrences scoring functions. Our analytics is optimized to allow efficient parallel data processing and produce a compact output dataset.

Summarizing relationships between chemicals, genes, proteins, and diseases in patents

Fall 2022 abstracts, posters, and presentations will launch in mid-September.

ACS Fall 2022

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.