Leveraging Large Language Model to Generate Multi-modal Timeline Summarization

In the current era of abundant information, the skill to extract crucial events and present them in a succinct manner holds significant importance. Multi-modal timeline summarization (MTS) is especially pivotal across various domains such as news reporting, historical analysis, and social media monitoring. The recent progress in natural language processing has led to the emergence of large language models as potent tools for generating accurate and coherent summaries. However, the application of Large Language Models (LLMs) in multi-modal summarization faces challenges, including the scarcity of extensive multi-modal datasets for training and evaluation, as well as the requirement for more sophisticated evaluation metrics that can accommodate the multi-modal nature of the task, in contrast to current metrics designed for text summarization. Addressing these challenges necessitates the development of larger and more diverse datasets, along with the creation of novel evaluation metrics tailored to multi-modal summarization. This study aims to investigate the utilization of large language models in generating timeline summarizations, elucidating their capabilities, challenges, and potential applications.