Semantic Error Correction and Decoding for Short Block Channel Codes

arXiv cs.AI / 4/27/2026

📰 NewsDeveloper Stack & InfrastructureIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper proposes a receiver framework that uses multiple short block codes to send ASCII-encoded natural-language sentences over noisy AWGN channels, then repairs decoding errors using semantic context from a language model.
It introduces semantic error correction (SEC) to reconstruct corrupted segments, semantic list decoding (SLD) to generate and select candidate reconstructions via weighted Hamming distance, and a semantic confidence-guided HARQ (SHARQ) method that replaces CRC-based detection with confidence scoring to enable selective retransmissions without CRC overhead.
All components are implemented and trained with BART-style bidirectional and auto-regressive transformers, integrating semantic understanding directly into the physical-layer decoding pipeline.
Simulations show substantial improvements over conventional short and long capacity-approaching codes at the same rate, with BLER gains of about 0.4 dB from SEC, 0.8 dB additional improvement from SLD, and about 1.5 dB extra gain from SHARQ.
Compared with sending the whole sentence as a single long 5G LDPC codeword, the method reportedly improves semantic fidelity and reduces decoding latency by up to 90% due to segment-wise parallel processing.

Abstract

This paper presents a semantic-enhanced receiver framework for transmitting natural language sentences over noisy wireless channels using multiple short block codes. After ASCII encoding, the sentence is divided into segments, each independently encoded with a short block code and transmitted over an AWGN channel. At the receiver, segments are decoded in parallel, followed by a semantic error correction (SEC) model, which reconstructs corrupted segments using language model context. We further propose the semantic list decoding (SLD), which generates multiple candidate reconstructions and selects the best one via weighted Hamming distance, and a semantic confidence-guided HARQ (SHARQ) mechanism that replaces CRC-based error detection with a confidence score, enabling selective segment retransmission without CRC overhead. All modules are designed and trained using bidirectional and auto-regressive transformers (BART). Simulation results demonstrate that the proposed scheme significantly outperforms conventional capacity-approaching short codes and long codes at the same rate. Specifically, SEC provides approximately 0.4 dB BLER gain over plain short-code transmission, while SLD extends this to 0.8 dB. Compared to transmitting the entire sentence as a single long 5G LDPC codeword, our approach significantly improves semantic fidelity and reduces decoding latency by up to 90\%. SHARQ further provides an additional 1.5 dB gain over conventional HARQ.