Semantic Communications - A Comprehensive Survey for Future Research Issues

Semantic communication, aimed at conveying data meaning, emerges as a promising approach within information theory to address the limitations of current communication systems. This paper outlines the architecture of semantic communication, focusing on recent trends in semantic source and channel coding, as well as resource allocation. We specifically investigate recent deep learning (DL)-based semantic communication methodologies, including end-to-end (E2E) autoencoder models and Transformer-based structures, and discuss semantic coding with various data types such as text, image, speech, and multimodal data. Additionally, we analyze adaptive channel coding and strategies for resource allocation techniques from the perspective of semantic efficiency in resource-limited networks. The paper aims to propose future research directions by illustrating how semantic communication effectively overcomes challenges in resource-limited environments while enhancing the efficiency and reliability of communication systems.