2025
AI and LLM in Urdu
Multilingual hate speech detection in social media using translation-based approaches with large language models
Authors:
M. Usman M. Ahmad M. S. Tash I. Gelbukh…
… of 10,193 tweets in English,Urdu, and Spanish, annotated with … filter before the transformerorLLMencoder block, enhancing the … , with implications for equitableAIdevelopment. Future …
2025
AI and LLM in Urdu
An Efficient Approach for Code-Mixed Emotion Classification applying Machine Learning
Authors:
A. Mahmood M. Torres-Ruiz Z. Ahmad H. Farid…
…Urdu, with 7,474 instances; (ii) RomanUrdu, where the SMS messages are entirely in RomanUrdu, … 4, a GenerativeAIsection is presented, whereLLMdefines the generativeAImodel …
2025
AI and LLM in Urdu
Beyond Specialization: Benchmarking LLMs for Transliteration of Indian Languages
Authors:
G. Azam M. Sadique S. Ali M. Nadeem…
… Therefore, for all languages except forUrdu(forUrdulangauge, AK-Freq subset was … In ourwork, fine-tuning refers to the process of further training aLLMfor a specific task, which in our …
2025
AI and LLM in Urdu
A Sentence‐Level Encoder–Decoder Architecture for Designing an Administrative RomanUrduChatbot
Authors:
M. N. Maqbool R. M. Saleem N. Sarwar…
… administrative chatbot for RomanUrduto overcome the … orartificialintelligence(AI) based,the core objective of a chatbot is to effectively address user queries across diverse domains.AI…
2025
AI and LLM in Urdu
ImprovingLLMAbilities in Idiomatic
Authors:
S. Donthi M. Spencer O. Patel J. Y. Doh…
… to theLLMevaluations previously favoring the usage of the figurative meaning in thetranslation rather than a corresponding idiom, which is especially true here because, for theUrdu…
2025
AI and LLM in Urdu
Multilingual Cyber Threat Detection in Tweets/X Using ML, DL, andLLM: A Comparative Analysis
Authors:
S. A. Murad A. Dahal N. Rahimi- I. E. E. E. Transactions On … 2025
… text into theirUrducorpus and then working on theUrdulanguage with theirAIimplementation.The authors fine-tuned RoBERTa with 1313 English and 2400Urdusamples. The …
2025
AI and LLM in Urdu
Towards robustUrduaspect-based sentiment analysis through weakly-supervised annotation framework
Authors:
Z. Maqsood S. Latif R. Latif- … Of The 8Th International Conference On … 2025
…Urdu’s critical resource gap through scalable dataset creation methodology that eliminatesmanual annotation bottleneck to facilitate fine-grainedUrdu…LLMin Figure 4 (see appendix). …
2025
AI and LLM in Urdu
Trends and Challenges in Authorship Analysis: A Review of ML, DL, andLLMApproaches
Authors:
N. Habib T. Adewumi M. Liwicki E. Barney- arXiv Preprint arXiv … 2025
… The rise ofAI-…UrduNews Authorship attribution Corpus (UNAAC-20). They tested it withvarious ML and DL models and reported that CNN is the most effective technique for theUrdu…
2025
AI and LLM in Urdu
UrduLLaMA 1.0: Dataset Curation, Preprocessing, and Evaluation in Low-Resource Settings
Authors:
L. Fiaz M. H. Tahir S. Shams S. Hussain- arXiv Preprint arXiv:2502.16961 2025
… forUrduand other underrepresented languages. In this research, we tackle this challenge bydeveloping anUrdu-specificLLM… 2024) on 128 millionUrdutokens to enhance the model’s …
2025
AI and LLM in Urdu
Ai-powered linguistics: the digital transformation of language and text analysis
Authors:
H. Abbas D. Ahmad A. Hasham…
…AI-driven speech recognition, contextual awareness, and language support, particularly forUrdu… These conclusions contribute to the broader discussion regarding the application ofAI…
2025
AI and LLM in Urdu
Part of speech (POS) tagging in RomanUrdu: Datasets and models
Authors:
A. Faheem F. Ullah U. Azam M. S. Ayub…
…Urdu. In this work, we created a comprehensive, large-scale RomanUrduPOS dataset anddeveloped a RomanUrdu… a comprehensive framework for RomanUrduPOS tagging that …
2025
AI and LLM in Urdu
MALT: Mechanistic Ablation of Lossy Translation in LLMs for a Low-Resource Language:Urdu
Authors:
T. S. Bajwa- arXiv Preprint arXiv:2502.00041 2025
… For machine translation of English outputs generated from the editedLLMintoUrdu, thereare …AIgenerated dataset used in our methodology does not contain any harmful content or …