The source codes and dataset for our EMNLP 2022 findings paper: PcMSP: A Dataset for Scientific Action Graphs Extraction from Polycrystalline Materials Synthesis Procedure Text.
https://arxiv.org/abs/2210.12401
Original folder contains annotated raw data; json files are processed data
Due to license limit, the data is slightly different from results reported in our paper: 242 rather than 243 training files, 30 rather than 31 files for test.
Link will be made to the public soon.