Although perhaps more widely known for his later work on bibliometrics, Eugene Garfield (1925-2017) is also acknowledged within the chemical text mining community as the first person to describe the entity recognition of chemical names (and chemical formulae) in free text, and their extraction as chemical formulae for indexing and search. What is now considered a branch of Natural Language Processing (NLP) and Artificial Intelligence (AI), used for mining the current Big Data of journal articles and patents, actually traces its origins back to Eugene's 1962 PhD thesis from the University of Pennsylvania's Department of Linguistics. Modern "name-to-structure" software and chemical text mining tools not only owe their ancestry to algorithms first described over 50 years ago, but the ability of those original approaches to semantically resolve chemicals, and to handle ambiguous or generic structures places them at what is considered the state-of-the-art even today. In this talk we describe recent advances in text mining of chemical formulae and explain why Eugene Garfield's pragmatic approach to indexing chemical documents is in many ways superior to the recent vogue (fad?) of applying deep learning using LSTM (long short-term memory) recurrent neural networks or conditional random fields (CRFs) to chemical text mining.

Eugene Garfield: The father of chemical text mining and Artificial Intelligence (AI) in cheminformatics

255th National Meeting (2018)

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