The SOFC-Exp Corpus and Neural Approaches to Information Extraction in the Materials Science Domain

This paper presents a new challenging information extraction task in the\ndomain of materials science. We develop an annotation scheme for marking\ninformation on experiments related to solid oxide fuel cells in scientific\npublications, such as involved materials and measurement conditions. With this\npaper, we publish our annotation guidelines, as well as our SOFC-Exp corpus\nconsisting of 45 open-access scholarly articles annotated by domain experts. A\ncorpus and an inter-annotator agreement study demonstrate the complexity of the\nsuggested named entity recognition and slot filling tasks as well as high\nannotation quality. We also present strong neural-network based models for a\nvariety of tasks that can be addressed on the basis of our new data set. On all\ntasks, using BERT embeddings leads to large performance gains, but with\nincreasing task complexity, adding a recurrent neural network on top seems\nbeneficial. Our models will serve as competitive baselines in future work, and\nanalysis of their performance highlights difficult cases when modeling the data\nand suggests promising research directions.\n