TRANSFORMER-BASED CLASSIFICATION OF EUROPEAN COURT OF HUMAN RIGHTS CASES

Abstract

Transformer-based models like BERT and its subordinates are famous for text arrangement. Memory and computational limitations limit their utilization for classifying long archives across areas. These models additionally battle in languageexplicit regions like lawful texts because of their pre-preparing on wide dialects. This review arranges authoritative records, fundamentally from the ECHR dataset, underlining the requirement for powerful lawful portrayal and robotized order to work on legitimate effectiveness and save costs.A sliding window procedure handles enormous texts, while transformer-based models and traditional ML are utilized for classification. The venture utilizes Transfer Learning with BERT, RoBERTa, BigBird, Electra, and XLNet to order authoritative archives, especially for common liberties situations.In expansion, outfit approaches like Voting Classifier and LSTM, LSTM + GRU models are utilized to further develop arrangement accuracy to 92%. For client testing and openness, a Flask-based frontend with verification is recommended. This exploration progresses mechanized authoritative record characterization, further developing lawful guide administration, cost decrease, and openness.