Empowering Prior to Court Legal Analysis: A Transparent and Accessible Dataset for Defensive Statement Classification and Interpretation

• DOI: 10.23919/fruct61870.2024.10516405
• Publication Date: 2024-05-17T17:20:55Z
• Authors: Yannis Spyridis; Jean-Paul Younes; Haneen Deeb; Vasileios Argyriou
• Keywords: Interpretability; Statement (logic)

The classification of statements provided by individuals during police interviews is a complex and significant task within the domain natural language processing (NLP) legal informatics. lack extensive domain-specific datasets raises challenges to advancement NLP methods in field. This paper aims address some present introducing novel dataset tailored for made interviews, prior court proceedings. Utilising curated training evaluation, we introduce fine-tuned DistilBERT model that achieves state-of-the-art performance distinguishing truthful from deceptive statements. To enhance interpretability, employ explainable artificial intelligence (XAI) offer explainability through saliency maps, interpret model's decision-making process. Lastly, an XAI interface empowers both professionals non-specialists interact with benefit our system. Our accuracy 86%, shown outperform custom transformer architecture comparative study. holistic approach advances accessibility, transparency, effectiveness statement analysis, promising implications practice research.