An Approach Toward Abstractive Text Summarization for Urdu Language Using LLM (ATSUL)

This paper explores the challenges and opportunities surrounding Urdu, a low-resourced language, particularly in the context of text summarization techniques and the utilization of large language models (LLMs). The paper begins with a comprehensive literature survey, exploring existing research and methodologies pertinent to Urdu language processing, summarization, and the application of LLMs. It delves into the unique characteristics of Urdu, such as its morphology, script, and linguistic complexities, which pose challenges for natural language processing tasks. Furthermore, the paper examines various abstractive text summarization techniques highlighting their applicability and limitations in the Urdu language domain. Additionally, it discusses the advent of large language models and their potential to address issues related to low-resource languages like Urdu, by leveraging vast amounts of data and sophisticated neural architectures. Through an in-depth analysis and synthesis of existing literature, this paper emphasizes on the processing of Urdu language, abstractive text summarization using MT0 (Multitask prompted Fine Tuning) using BBC news dataset.