Sualeha Farid

Samee Arif, Sualeha Farid, Awais Athar et al.

This paper introduces UQA, a novel dataset for question answering and text comprehension in Urdu, a low-resource language with over 70 million native speakers. UQA is generated by translating the Stanford Question Answering Dataset (SQuAD2.0), a large-scale English QA dataset, using a technique called EATS (Enclose to Anchor, Translate, Seek), which preserves the answer spans in the translated context paragraphs. The paper describes the process of selecting and evaluating the best translation model among two candidates: Google Translator and Seamless M4T. The paper also benchmarks several state-of-the-art multilingual QA models on UQA, including mBERT, XLM-RoBERTa, and mT5, and reports promising results. For XLM-RoBERTa-XL, we have an F1 score of 85.99 and 74.56 EM. UQA is a valuable resource for developing and testing multilingual NLP systems for Urdu and for enhancing the cross-lingual transferability of existing models. Further, the paper demonstrates the effectiveness of EATS for creating high-quality datasets for other languages and domains. The UQA dataset and the code are publicly available at www.github.com/sameearif/UQA.

Samee Arif, Sualeha Farid

This paper introduces an end-to-end pipeline for Optical Character Recognition (OCR) on Urdu newspapers, addressing challenges posed by complex multi-column layouts, low-resolution scans, and the stylistic variability of the Nastaliq script. Our system comprises four modules: (1) article segmentation, (2) image super-resolution, (3) column segmentation, and (4) text recognition. We fine-tune YOLOv11x for segmentation, achieving 0.963 precision for articles and 0.970 for columns. A SwinIR-based super-resolution model boosts LLM text recognition accuracy by 25-70%. We also introduce the Urdu Newspaper Benchmark (UNB), a manually annotated dataset for Urdu OCR. Using UNB and the OpenITI corpus, we compare traditional CNN+RNN-based OCR models with modern LLMs. Gemini-2.5-Pro achieves the best performance with a WER of 0.133. We further analyze LLM outputs via insertion, deletion, and substitution error breakdowns, as well as character-level confusion analysis. Finally, we show that fine-tuning on just 500 samples yields a 6.13% WER improvement, highlighting the adaptability of LLMs for Urdu OCR.