Evaluation of Automatic Legal Text Summarization Techniques for Greek Case Law

The increasing amount of legal information available online is overwhelming for both citizens and professionals, making it difficult time-consuming to find relevant keep up with the latest developments. Automatic text summarization techniques can be highly beneficial as they save time, reduce costs, lessen cognitive load professionals. However, applying these documents poses several challenges due complexity lack needed resources, especially in linguistically under-resourced languages, such Greek language. In this paper, we address automatic documents. A major challenge area suitable datasets response, developed a new metadata-rich dataset consisting selected judgments from Supreme Civil Criminal Court Greece, alongside their reference summaries category tags, tailored purpose automated document summarization. We also adopted state-of-the-art methods abstractive extractive conducted comprehensive evaluation using human metrics. Our results: (i) revealed that, while exhibit average performance, generate moderately fluent coherent text, but tend receive low scores relevance consistency metrics; (ii) indicated need metrics that capture better summary’s coherence, relevance, consistency; (iii) demonstrated fine-tuning BERT models on specific upstream task significantly improve model’s performance.