THE SYNTHESIS PROJECT
A Data-Driven Framework for Materials Synthesis Discovery
ABOUT THE PROJECT
Advances in computational materials design have enabled rapid screening for desirable properties of both real and virtual compounds. However, the pace of commercially-realized advanced materials may now be limited by trial-and-error synthesis techniques.
The goal of this research project is to advance computational learning around materials synthesis approaches by creating a predictive synthesis system for advanced materials design and processing — to do for materials synthesis what modern computational methods have done for materials properties.
This predictive synthesis system has been applied to several materials science domains including metallic oxides, titania, perovskites, zeolites, metal-insulator transition materials, solid-state electrolytes, alloys, and alternative cements.
MATERIALS WORD EMBEDDINGS
Get pre-trained word vectors for materials science. Word2Vec, FastText, and ELMO embeddings available. A useful starting point for text-mining!
NATURAL LANGUAGE PROCESSING MODELS
Get pre-trained classifiers to identify synthesis sections of paper and perform synthesis-relevant named entity recognition.
TABLE DATA EXTRACTOR
Software to automatically extract data from tables embedded in HTML and XML files in JSON structures.
EXTRACTED DATA SETS
Various public materials data sets extracted with the Olivetti group NLP pipeline.
This project is closely aligned with the Materials Genome Initiative (MGI) for Global Competitiveness, a multi-agency effort for deploying advanced materials. We are supported by funding from the National Science Foundation Award DMREF Awards #1922311, #1922372, and #1922090, the Office of Naval Research (ONR) under Contract No. N00014-20-1-2280, and the MIT Energy Initiative. Early work on the project was also collaborative under the Department of Energy’s Basic Energy Science Program through the Materials Project under Grant No. EDCBEE.
More information about the project and related research can be found at http://olivetti.mit.edu/.