HOW TO RUN ?
-
For generating the definitions: 1.1)run : python src\generate_defs_corpus.py It reads config from .\config\config_defs.yaml The source file is '.\data\all_sources_metadata_2020-03-13.csv.' The extraction is done from sciScpacy model. The extraction itself is time consuming , hence fr=or each iteration it saves the defintion at temp location and later run the script mentioned in 1.2 to get the final output.
For sciSpcay set-up use this link: https://allenai.github.io/scispacy/
1.2) run: python src\create_definition_file.py It reads config from .\config\config_defs.yaml and generates output at .\data\definition-corona-final.csv
-
For semantic role tagging using BERT
run: python src\ser_bert_tagger.py --mode train\eval This implementation uses BERT to produce the ser tags. It reads config from ./config/config_defs.yaml . Read about BERT here : https://github.com/google-research/bert
Download the trained model from here : https://drive.google.com/open?id=1p_sb3yxr_n6IBfs2QqlKKIO2Ygk5GBNk
NOTE : You may need to unzip the files in .\data and place directly under .\data folder for processing or you can download the data and model from the link shared above.
This implementation usage tensorflow version 1.15 and python 3.7. Download all requirements pip install -r requirements.txt