The classification of medieval Hebrew literature has long relied on historically inherited genre labels, often leading to misassignments and blurred textual boundaries. This study applies transformer-based cluster analysis to a corpus of 60 texts, using BEREL embeddings and hierarchical clustering to evaluate whether selected computational methods can provide empirically grounded, complementary insights to traditional genre classifications. The analysis identifies several stable clusters, including a distinct Narrative cluster, reinforcing prior research that questions the categorization of certain texts as Aggadic Midrash. While some clusters align with established classifications, others highlight ambiguities that challenge conventional taxonomies. The study demonstrates that computational clustering can systematically capture textual affinities, revealing relationships that may remain obscured in traditional approaches. These findings establish a methodological framework for reassessing genre structures in Hebrew literature, laying the groundwork for future research based on expanded datasets and manuscript evidence.
Annabelle Fuchs (Fri,) studied this question.