Deciphering the Underserved: Benchmarking LLM OCR for Low-Resource Scripts

This study investigates the potential of Large Language Models (LLMs), particularly GPT-4o, for Optical Character Recognition (OCR) in low-resource scripts such as Urdu, Albanian, and Tajik, with English serving as a benchmark. Using a meticulously curated dataset of 2,520 images incorporating controlled variations in text length, font size, background color, and blur, the research simulates diverse real-world challenges. Results emphasize the limitations of zero-shot LLM-based OCR, particularly for linguistically complex scripts, highlighting the need for annotated datasets and fine-tuned models. This work underscores the urgency of addressing accessibility gaps in text digitization, paving the way for inclusive and robust OCR solutions for underserved languages.

Quantitative Ink Analysis: Estimating the Number of Inks in Documents through Hyperspectral Imaging

In the field of document forensics, ink analysis plays a crucial role in determining the authenticity of legal and historic documents and detecting forgery. Visual examination alone is insufficient for distinguishing visually similar inks, necessitating the use of advanced scientific techniques. This paper proposes an ink analysis technique based on hyperspectral imaging, which enables the examination of documents in hundreds of narrowly spaced spectral bands, revealing hidden details. The main objective of this study is to identify the number of distinct inks used in a document. Three clustering algorithms, namely k-means, Agglomerative, and c-means, are employed to estimate the number of inks present. The methodology involves data extraction, ink pixel segmentation, and ink number determination. The results demonstrate the effectiveness of the proposed technique in identifying ink clusters and distinguishing between different inks. The analysis of a hyperspectral cube dataset reveals variations in spectral reflectance across different bands and distinct spectral responses among the 12 lines, indicating the presence of multiple inks. The clustering algorithms successfully identify ink clusters, with k-means clustering showing superior classification performance. These findings contribute to the development of reliable methodologies for ink analysis using hyperspectral imaging, enhancing the