This is the repository for the paper "Extracting Space Situational Awareness Events from News Text" (Xie et al., Under Review).
The labelled spans can be found in the data folder, and are presened in the CoNLL BIO format.
The performance of the span labelling system can be found in results.
Preditions of the system on the test set can be found in results.
Subsequent analyses (including micro-averages) can be found in results/analysis.
The open source span labelling system can be found at: https://github.com/kamalkraj/BERT-NER , which requires customization using the code found in this repository in the code folder.
For training, replace run_ner.py
with the version customized for this SSA corpus. The included runme_60.sh
script will complete the training and evaluation procedure for each of the three SSA events (launches, failures, and decommissionings). For inference (label prediction), use inference.py
.
Should you have any questions, comments, or issues, please feel free to get in touch at pajansen@arizona.edu
.