# INCANT
## Raw Data
Raw data used to generate INCANT networks is scraped from GDELT using `Final_Datasets/data_pull.py`. The output for each domain is stored in `Final_Datasets/[domain name].csv`.
## Network Generation
INCANT networks are generated from raw data using `process.py`. The output networks are stored in `Final_Datasets/[domain name]_network.csv`. These csv files contain columns for source and target nodes, as well as additional metadata.
Helper functions for the QA task and coreference resolution used in network generation are found under the `QA` and `COREF` directories respectively. Template questions for QA are found in `template.py`.
# TAMPA
## Networks
Input networks to TAMPA generated by INCANT are found in the `input/` directory. The contents of this directory consist of the 6 domains analyzed in the paper. The outputs of message passing on these INCANT networks are found in the `output/` directory. This directory contains outputs of message passing on both the INCANT and PERM networks.
## Message Passing and Clustering
Functions to perform message passing on the INCANT networks are found in `msg_passing.py`. Examples of using these can be found in `run.py`. Functions to perform clustering on node embeddings can be found in `parse_data.py`. Helpful functions for visualizing results can be found in `display.py`. Examples of using these can be found in `msg_passing.ipynb`.
## Baselines
Functions for both baseline models are found in `baselines.py`.