Papers/2608.19385
🧪 Test?View on arXiv

Beyond Recognition: Compact Multi-Domain Arabic Manuscript HTR with Candidate-Selection Analysis and Evidence-Preserving Review

Not provided in the content

HTRevidence managementmulti-domainArabic manuscripts
2608.19385
Builder Relevance
80%
2h ago

Abstract

The paper presents a system for historical Arabic manuscript transcription that addresses recognition challenges and incorporates an evidence-aware review process.

Reality Card

Core Claim

Phoenix, a CNN-BiLSTM-CTC recognizer, achieved a character error rate (CER) reduction from 22.12% to 17.86% on Agapet lines and from 17.72% to 11.84% on Omar lines, demonstrating significant improvements in multi-domain manuscript transcription.

Method / Result

Character-weighted CER fell from 19.98% to 14.93%, a 25.3% relative error reduction.

Limitations

The system's performance may regress in certain domains, as indicated by the increase in CER on TariMa lines.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers