DisContNet: Contradiction Detection in Texts using Transformers

In this study, the significance of contradiction detection in natural language text is highlighted, and the study proposes a comprehensive system that leverages advanced natural language processing (NLP) and deep learning methodologies. The approach, constructed around a dataset derived from accessible corpora, aims to classify pairs of sentences based on the type of contradiction present within them. The system not only detects contradictions but also distinguishes among various types of contradictions. Further, the performances of two kinds of Transformer Models (BERT and DistilBERT) are compared. The findings of this research pave the way for further enhancement of NLP techniques for contradiction detection, demonstrating its potential applicability across diverse NLP tasks.