MatSciBERT software results when applied to a magnetic paper corpus

This document contains the analyse of the results when MatSciBERT software was applied to the magnetic paper corpus, as part of the "Feasibility study to assess Natural Language Processing for Functional Magnetic Materials" project funded by ROyce Materials 4.0. As MatSciBERT gives a label to every word with the abstract, this means that the output for the data is not as simple as the other two data sets. Rather the raw data output is 100’s of lines long. Thus the raw data output for the first corpus is found here: here The analysed data is given in the document, along with the highlighted expected results and a comment on how well the software worked.