Understanding the Impact of Evidence-Aware Sentence Selection for Fact Checking
Implementation of the papers Understanding the Impact of Evidence-Aware Sentence Selection for Fact Checking and A Review on Fact Extraction and Verification.
Requirements
See requirements.txt
Task
The goal is given a claim and a recent dump of Wikipedia documents to predict the veracity of the claim.
Run
Angular loss
For the retrieval step in the ranking_nlp4if/angular_embedding/ directory run:
bash pipeline.sh
For the classification in the ranking_nlp4if/kgat/ directory run:
bash pipeline_angular_kgat.sh
Evidence-aware loss
For the retrieval step in the ranking_nlp4if/pointwise_transformer/ directory run:
bash pipeline_evi_num_5_slate_20.sh
For the classification in the ranking_nlp4if/kgat/ directory run:
bash pipeline_transformer_evi_num_5_slate_20_kgat.sh
Cosine loss
For the retrieval step in the ranking_nlp4if/cosine_classifier/ directory run:
bash pipeline.sh
For the classification in the ranking_nlp4if/kgat/ directory run:
bash pipeline_kgat_cosine.sh
Pairwise loss
For the retrieval step in the ranking_nlp4if/pairwise/ directory run:
bash pipeline_pairwise.sh
For the classification in the ranking_nlp4if/kgat/ directory run:
bash pipeline_pairwise_kgat.sh
Pointwise loss
For the retrieval step in the ranking_nlp4if/pointwise/ directory run:
bash pipeline.sh
For the classification in the ranking_nlp4if/kgat/ directory run:
bash pipeline_pointwise_kgat.sh
Acknowledgement
Code and preprocessed data for the experiments on the FEVER dataset are adapted from the KernelGAT repository.
Code for the evidence-aware model is based on the allRank repository.
Notes
Please cite our work when using this software.
Bekoulis, G., Papagiannopoulou, C., & Deligiannis, N. (2021, June). Understanding the Impact of Evidence-Aware Sentence Selection for Fact Checking. In Proceedings of the Fourth Workshop on NLP for Internet Freedom: Censorship, Disinformation, and Propaganda (pp. 23-28).
Bekoulis, G., Papagiannopoulou, C., & Deligiannis, N. (2020). Fact Extraction and VERification--The FEVER case: An Overview. arXiv preprint arXiv:2010.03001.