Tathagato Roy supervised by Dr. Rahul Mishra received his Master of Science – Dual Degree  in Computer Science and Engineering (CSD). Here’s a summary of his research work on Toward Multi Attribute Controllable Summarization:

 Text summarization has long been a key task in natural language processing (NLP), but traditional methods often fail to align with the specific needs and goals of individual users. Recent research has shifted towards controllable summarization techniques that better address these unique user requirements. However, despite growing interest in controllable summarization, a comprehensive survey that thoroughly explores the various controllable attributes, the challenges they pose, and the solutions developed is still lacking. This thesis first formally defines the Controllable Text Summarization (CTS) task, categorizes controllable attributes based on shared characteristics and objectives, and provides an in-depth review of existing datasets and methods within each category. We also identify key limitations and research gaps in the field and explore potential solutions and future research directions. While progress has been made in controllable summarization, the investigation of Multi-Attribute Controllable Summarization (MACS) is still underdeveloped. This paper aims to fill this gap by analyzing the MACS task through the use of large language models (LLMs) and various learning paradigms, with a focus on low-rank adapters. We experiment with different fine-tuning strategies to evaluate the effectiveness of models in retaining patterns associated with multiple controllable attributes. Additionally, we propose a novel hierarchical adapter fusion technique to integrate knowledge from two distinct controllable attributes. Our findings, challenges, and suggestions for future advancements in the MACS task are presented.

June 2025