Publications/Natural Language Processing & LLMs for Low-Resource Languages
Conference · 2024
UQA: Corpus for Urdu Question Answering
Samee Arif, Sualeha Farid, Awais Athar, Agha Ali Raza
LREC-COLING 2024, Torino, Italy
Summary
Question-answering systems need large training sets, which Urdu lacked. UQA is created by translating the Stanford Question Answering Dataset (SQuAD2.0) into Urdu with a technique we call EATS (Enclose to Anchor, Translate, Seek), which preserves the position of each answer inside its translated paragraph. We select the better of two candidate translation systems and benchmark multilingual models on the result; the best, XLM-RoBERTa-XL, achieves an F1 score of 85.99. The dataset and code are public, and the same technique can create question-answering data for other languages.
Cite this paper
Samee Arif, Sualeha Farid, Awais Athar, Agha Ali Raza. UQA: Corpus for Urdu Question Answering. LREC-COLING 2024, Torino, Italy.
@inproceedings{arif2024uqa,
author = {Samee Arif and Sualeha Farid and Awais Athar and Agha Ali Raza},
title = {{UQA: Corpus for Urdu Question Answering}},
booktitle = {LREC-COLING 2024, Torino, Italy},
year = {2024},
url = {https://aghaaliraza.com/publications/2024--LREC-COLING--UQA_Corpus_for_Urdu_Question_Answering.pdf}
}
- IRTIQA: Evolving in English, Evaluated in Urdu — Does Agent Self-Evolution Transfer Across Languages?Findings of EACL 2027, Athens, Greece
- UrduCultureQA: A Culturally Grounded Benchmark for Urdu Question Answering and Alignment EvaluationFindings of EACL 2027, Athens, Greece
- Pak3H: Evaluating the Cost of Cultural Mismatch in LLM Alignment with a Human-Contextualized Urdu BenchmarkEMNLP 2026 (Main), Budapest, Hungary
- Evaluating Robustness of LLM-Based Ambient Scribes for SOAP Note GenerationCHIL 2026 (Conference on Health, Inference, and Learning), PMLR vol. 333
- Kahaani: A Multimodal Co-Creative Storytelling SystemEACL 2026 Student Research Workshop, Rabat, Morocco