Semantic communication shifts the design objective of communication systems from exact bit reconstruction toward the preservation of meaning and receiver-side utility. This shift is particularly critical in scenarios where small bit-level perturbations can alter numbers, negations, or operational commands, transforming a message while maintaining high lexical similarity. This paper proposes a Bit Error Rate (BER)-driven semantic resilience framework for the controlled evaluation of text-based communication. The framework models the transmission pipeline from binary encoding to channel corruption, decoding, and optional receiver-side semantic recovery using Large Language Models (LLMs). We introduce a multi-level evaluation methodology combining Word Error Rate (WER), embedding-based Semantic Textual Similarity (STS), and Semantic Textual Similarity Gain (STS Gain). Unlike conventional purely bit-oriented approaches, the proposed protocol explicitly evaluates the trade-off between exact lexical reconstruction and semantic robustness. The results provide a reproducible testbed for benchmarking semantic recovery strategies under varying BER conditions.

Semantic Communication,Text Transmission,Bit Error Rate,Receiver-Side Recovery,Word Error Rate,Sentence-BERT,6G