MULTIMODAL AI FOR ENHANCED CONTENT ACCESSIBILITY: A SURVEY ON TEXT, SPEECH, AND IMAGE PROCESSING

ABSTRACTThis survey explores the role of multimodal artificial intelligence (AI) in processing text, speech, and images, focusing on Multimodal Large Language Models (MLLMs). MLLMs integrate various data types, such as text, audio, and images, advancing fields like natural language processing (NLP), computer vision, and speech recognition. The survey discusses the architecture, training methods, and data requirements of MLLMs, with an emphasis on transformer networks that process informationacross modalities. It highlights key applications like image captioning, cross-modal retrieval, real-time speech recognition, and content creation. The review also addresses challenges such as bias, inaccuracies, the need for large datasets, and computational demands, while examining the potential and future directions of MLLMs in AI.