openie

Assumption: The data is already downloaded using the steps mentioned in the Readme of the root directory.

Approach I: Stanford OpenIE

Run the stanford openie on the corpus and QA Hypothesis. Follow the steps mentioned here. You should have two dumps after this

1. path_to_corpus_triplets: Stanford OpenIE triplets of the entire corpus.
2. path_to_qa_hypo_graph_dump: Dump of hypothesis graphs.

Run the following command to create the corpus graph and use Approach I to answer these questions to save the predictions in path_to_pred_file.

Note: The corpus graph is saved for future use.

python run.py path_to_corpus_triplets path_to_qa_hypo_graph_dump \
    path_to_pred_file

Using the official evaluation script by ARC

python ../evaluator.py \
    -qa ../data/ARC-V1-Feb2018-2/ARC-Challenge/ARC-Challenge-Test.jsonl \
    -p path_to_pred_file \
    -o output_file

Approach I: StuffIE

Follow the steps in stuffie's Readme to Run stuffIE on corpus and convert the obtained Triplets into stanford's format.

Now you can use the above mentioned converted corpus instead of openIE.