Baybars: Arabic Historical Manuscript Line Images with Transcriptions

BAYBARS is a specialized dataset for Handwritten Text Recognition (HTR) focusing on Arabic historical manuscripts. It contains a curated selection of pages from the Sīrat Baybars (the Epic of Baibars) transcribed in the Oriental script (Levantine Middle Arabic). The dataset was created by calfa-ai and was last updated on HuggingFace in May 2026.

Use Cases

• Train Handwritten Text Recognition (HTR) models based on line-level images of Arabic script.
• Benchmark OCR performance on historical Arabic manuscripts based on the paired image-text data.
• Study linguistic features of Levantine Middle Arabic based on the transcribed text from the Sīrat Baybars.

Strengths

• Focuses on a specific, historically significant text (Sīrat Baybars).
• Provides curated, line-level image-text pairs for a specialized HTR task.
• Last updated on 2026-05-22, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect temporal and linguistic bias inherent to its single historical source.

Provenance

Source

calfa-ai on HuggingFace.

Collection Method

Likely involves curation and transcription of pages from the Sīrat Baybars manuscript.

Freshness

Last updated 2026-05-22 09:10:08; freshness should be verified.