Our Research

Abstract: 
As part of the CheckThat! 2020 Task 2, we investigated sentence similarity using transformer models. In Task 2, the goal was to effectively rank claims based on their relevancy compared to an input tweet. While setting our baseline on sentence similarity for fact-checking, we gathered insights we felt compelled to share in this paper. We learned how multi-modal data utilization could foster significant uplifts in model performance. We also gained knowledge on which hybrid training and strong sampling worked best for fact-checking applications, and wanted to share our interpretation of the results we got. Finally, we want to explain our recommendations on data augmentations. All of the above allowed us to set our baseline in fact-checking in the CLEF Checkthat! 2020 Task 2 competition.

Abstract: 
As part of the FEVEROUS shared task, we developed a robust and finely tuned architecture to handle the joint retrieval and entailment on text data as well as structured data like tables. We proposed two training schemes to tackle the hurdles inherent to multi-hop multi modal datasets. The first one allows having a robust retrieval of full evidence sets, while the second one enables entailment to take full advantage of noisy evidence inputs. In addition, our work has revealed important insights and potential avenue of research for future improvement on this kind of dataset. In preliminary evaluation on the FEVEROUS shared task test set, our system achieves 0.271 FEVEROUS score, with 0.4258 evidence recall and 0.5607 entailment accuracy.