Unified Language of Synthesis Actions for Representation of Synthesis Protocols—Making Steps Toward Autonomous Materials Synthesis

Over the past decade, we have witnessed growing success of data-driven and artificial intelligence (AI)-based methodologies promoting breakthroughs in predicting materials structure, properties, and functionality. Nonetheless, adapting AI power to predict and control materials synthesis is still challenging and requires substantial effort in gathering high-quality large-scale datasets. A possible approach would be to extract synthesis information from scientific publications. However, this route is still challenging, especially for extracting synthesis actions, because of the lack of a comprehensive labeled dataset using a solid, robust, and well-established ontology for describing synthesis procedures. Taking a step further, the growing interest in automated AI-guided materials synthesis demands a robust and unified language for describing synthesis protocols in order to make them applicable to the autonomous robotic platform.<br/>In this work, we propose a <i>unified language of synthesis actions </i>(ULSA) to describe solid-state, sol-gel, precipitation, and solvo-/hydrothermal synthesis procedures. We also present a labeled dataset of 3,040 synthesis sentences created using the proposed ULSA schema. To verify the applicability of the ULSA and the dataset, we trained a neural network-based model that identifies a sequence of synthesis actions in a paragraph, maps them into the ULSA, and builds a graph of the synthesis procedure. Analysis of the graphs from thousands of paragraphs has shown that this ULSA vocabulary is enough to obtain high-accuracy extraction of synthesis actions as well as to pick the important features of each of the aforementioned synthesis types. We anticipate that these results will be widely used by the researchers interested in scientific text mining and will help to achieve a breakthrough in predictive and AI-guided autonomous materials synthesis.