Superconductor Research Papers Clustering using Weighted Annotated Information

Abstract

We report an on-going work aiming to utilize artificial intelligence principles to support superconducting materials research. In particular, we want to facilitate relevant information access for superconducting materials researchers using automatic clustering. We use a weighted clustering schema for different categories of superconducting materials information (such as the class of the superconducting material, the critical temperature, or measurement method) to find similar research papers that discuss information category of interest. These information categories were extracted from the SuperMat corpus. We developed this corpus consisting of research papers annotated with linked 6 information categories related to superconductors development. We demonstrate that clustering research papers using the general content of the paper might not be efficient for researches interested in a specific information category. Instead, the weighted clustering schema can improve the clustering quality given a desired category of interest.