Coarse-to-Fine Semantic Communication Systems for Text Transmission

Achieving more powerful semantic representations and semantic understanding is one of the key problems in improving the performance of semantic communication systems. This work focuses on enhancing the semantic understanding of the text data to improve the effectiveness of semantic exchange. We propose a novel semantic communication system for text transmission, in which the semantic understanding is enhanced by coarse-to-fine processing. Especially, a dual attention mechanism is proposed to capture both the coarse and fine semantic information. Numerical experiments show the proposed system outperforms the benchmarks in terms of bilingual evaluation, sentence similarity, and robustness under various channel conditions.