Researchers have developed Phoenix, a 4.99-million-parameter CNN-BiLSTM-CTC model for transcribing historical Arabic manuscripts. This model is adapted across different manuscript domains using document-aware replay and specialized guards to maintain performance across new datasets. The system achieved a significant reduction in character error rate (CER) on large held-out sets, improving from 19.98% to 14.93%. Accompanying Phoenix is Athar, a review workflow designed to manage uncertain readings and preserve visual evidence, exporting auditable TEI and PAGE-XML records. AI
IMPACT Advances the state-of-the-art in historical document transcription, enabling more accurate scholarly analysis of Arabic manuscripts.
RANK_REASON The cluster contains an academic paper detailing a new model and workflow for a specific AI task. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
