Abstract
Experimental procedures for chemical synthesis are commonly reported in prose in patents or in the scientific literature. The extraction of the details necessary to reproduce and validate a synthesis in a chemical laboratory is often a tedious task requiring extensive human intervention. We present a method to convert unstructured experimental procedures written in English to structured synthetic steps (action sequences) reflecting all the operations needed to successfully conduct the corresponding chemical reactions. To achieve this, we design a set of synthesis actions with predefined properties and a deep-learning sequence to sequence model based on the transformer architecture to convert experimental procedures to action sequences. The model is pretrained on vast amounts of data generated automatically with a custom rule-based natural language processing approach and refined on manually annotated samples. Predictions on our test set result in a perfect (100%) match of the action sequence for 60.8% of sentences, a 90% match for 71.3% of sentences, and a 75% match for 82.4% of sentences.
Highlights
Experimental procedures for chemical synthesis are commonly reported in prose in patents or in the scientific literature
Some simple organic reaction data are widely presented in well-structured and machine readable format, this is not the case for the corresponding chemical reaction procedures which are reported in prose in patents and in scientific literature
The following is an example of a typical experimental procedure that is to be converted to automationfriendly instructions: To a suspension of methyl 3-7-amino-2-[(2,4-dichlorophenyl)methyl]-1H-benzimidazol-1-ylpropanoate (6.00 g, 14.7 mmol) and acetic acid (7.4 mL) in methanol (147 mL) was added acetaldehyde (4.95 mL, 88.2 mmol) at 0 ∘C
Summary
Experimental procedures for chemical synthesis are commonly reported in prose in patents or in the scientific literature. Some simple organic reaction data are widely presented in well-structured and machine readable format, this is not the case for the corresponding chemical reaction procedures which are reported in prose in patents and in scientific literature. It is not surprising if their conversion into a structured format is still a daunting task. The design of an automated conversion from unstructured chemical recipes for organic synthesis into structured ones is a desirable and needed technology With such an algorithm, a machine could ingest an experimental procedure and automatically start the synthesis in the lab, provided that all the necessary chemicals are available. If applied to a large collection of experimental procedures, the conversion to structured synthesis actions could prove interesting for the analysis of reaction data, and could facilitate the discovery of patterns and the training of machine-learning models for new organic chemistry applications
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
