Low-Resource Heuristics for Bahnaric Optical Character Recognition Improvement

TL;DR

The paper tackles the challenge of digitizing Bahnar language documents with scarce data by designing a hybrid OCR pipeline that combines layout-aware table detection with probabilistic post-processing. It employs traditional computer vision for structure analysis (via OpenCV and Hough transforms) and a Bahnar-specific n-gram dictionary to drive a sliding-window error-correction heuristic, avoiding data-hungry neural models. The approach yields a notable improvement in recognition accuracy, achieving $79.26\%$ on representative samples from $72.86\%$, and demonstrates the viability of resource-efficient digitization for minority languages. By delivering a Bahnar-focused dataset and a reproducible processing framework, the work advances language preservation, lexicography, and NLP research for low-resource languages and can be extended to similar orthographies and table-rich lexical resources.

Abstract

Bahnar, a minority language spoken across Vietnam, Cambodia, and Laos, faces significant preservation challenges due to limited research and data availability. This study addresses the critical need for accurate digitization of Bahnar language documents through optical character recognition (OCR) technology. Digitizing scanned paper documents poses significant challenges, as degraded image quality from broken or blurred areas introduces considerable OCR errors that compromise information retrieval systems. We propose a comprehensive approach combining advanced table and non-table detection techniques with probability-based post-processing heuristics to enhance recognition accuracy. Our method first applies detection algorithms to improve input data quality, then employs probabilistic error correction on OCR output. Experimental results indicate a substantial improvement, with recognition accuracy increasing from 72.86% to 79.26%. This work contributes valuable resources for Bahnar language preservation and provides a framework applicable to other minority language digitization efforts.

Figures (8)

• Figure 1: Original images of Bahnar Dialect Dictionary.
• Figure 2: Cropped images of Bahnar Dialect Dictionary.
• Figure 3: Cell corner identification via point-spreading over line intersections.
• Figure 4: Language model-enhanced OCR output for pages 5 and 6.
• Figure 5: Architecture for Bahnar text extraction from images.