Essential Step Toward Mining Big Polymer Data: PolyName2Structure, Mapping Polymer Names to Structures

Abstract

Advances in polymer science have made polymers essential in our everyday life and have yielded unprecedented quantities of data over the past several decades. However, it is still challenging and inefficient to organize such scattered and accumulated “big data” in a text format through mass journals, patents, and web pages due to the complexity and ambiguity of polymer representations. In this paper, we report the first automated framework, PolyName2Structure (PN2S), which is able to convert various polymer name representations to their corresponding polymer structures. In PN2S, machine learning models were built to predict the polymerization pathway, identify the reacting group(s), and generate repeating units after polymerization. This PN2S system achieved over 90% accuracy when applied to polymer names listed in a commercial catalog, embodying the first step toward resolving the complexity of the data structure for polymers by building a practical model that enables text mining of structural polymer information.