RerrFact
This repository contains the code for: RerrFact model for SciVer shared task.
Setup for Inference
1. Download SciFact database
Download the SciFact database from here.
2. Installing requirements
Install the requirements using the following command for abstract retrieval and rationale selection module.
pip install -r abstract,rationale_requirements.txt
Install the requirements using the following command for label prediction module.
pip install -r label_requirements.txt
3. Download trained models
Download the trained models using this link.
4. Using pre-trained models
Abstract Retrieval
python ./inference/abstract-retrieval.py \
--corpus ./data/corpus.jsonl \
--dataset ./data/claims_test.jsonl \
--model ./saved_models/abstract_retrieval_model_here \
--output ./prediction/abstract_retrieval_test_predictions.jsonl
Rationale Selection
python ./inference/rationale-selection.py \
--corpus ./data/corpus.jsonl \
--dataset ./data/claims_test.jsonl \
--abstract ./prediction/abstract_retrieval_test_predictions.jsonl \
--model ./saved_models/rationale_selection_model_here \
--output ./prediction/
Label Prediction
python inference/label-prediction.py \
--corpus '/data/corpus.jsonl' \
--dataset './data/claims_test.jsonl' \
--rationale-selection './prediction/rationale_selection.jsonl' \
--model_n './saved_models/neutral_classifer_here' \
--model_s './saved_models/support_classifier_here' \
--output './prediction/label_pred_test.jsonl'
Retrain models
Refer to training/Abstract-retrieval.ipynb for training abstract retrieval module.
Refer to training/Rationale-selection.ipynb for training rationale selection module.
Refer to training/Label-prediction.ipynb for training label prediction module.
1 Nov 19, 2021
47 Nov 28, 2022
4 Sep 26, 2022
61 Jan 07, 2023
328 Dec 26, 2022
20 Sep 12, 2021
9 Mar 03, 2022
3 Feb 11, 2022
236 Dec 26, 2022
193 Dec 15, 2022
15 Dec 17, 2022
85 Dec 14, 2022
110 Dec 03, 2022
251 Jan 08, 2023
2 Oct 06, 2022
45 Nov 21, 2022
744 Jan 04, 2023
6 Sep 28, 2022
35 Nov 29, 2022
2 Jun 21, 2022