Semantic Communications via Features Identification

TL;DR

This work addresses semantic communications by proposing Identification via semantic features, a framework that couples a teacher-apprentice architecture with a message-identification paradigm to exploit semantic ambiguity. The method maps learnable content to identities in a feature space and transmits selected features to progressively identify a semantic element, balancing accuracy and bit efficiency through a confidence threshold $\lambda$. Key contributions include formalizing a semantic base with SeB, integrating learnable/memorizable data streams, and demonstrating significant bit savings (e.g., ~18%–50% of syntactic bits) while preserving meaningful semantic identification in simulations. The approach promises practical bandwidth reductions for next-generation wireless networks, with avenues for refining feature selection and identification strategies.

Abstract

The development of the new generation of wireless technologies (6G) has led to an increased interest in semantic communication. Thanks also to recent developments in artificial intelligence and communication technologies, researchers in this field have defined new communication paradigms that go beyond those of syntactic communication to post-Shannon and semantic communication. However, there is still need to define a clear and practical framework for semantic communication, as well as an effective structure of semantic elements that can be used in it. The aim of this work is to bridge the gap between two post-Shannon communication paradigms, and to define a robust and effective semantic communication strategy that focuses on a dedicated semantic element that can be easily derived from any type of message. Our work will take form as an innovative communication method called identification via semantic features, which aims at exploiting the ambiguities present in semantic messages, allowing for their identification instead of reproducing them bit by bit. Our approach has been tested through numerical simulations using a combination of machine learning and data analysis. The proposed communication method showed promising results, demonstrating a clear and significant gain over traditional syntactic communication paradigms.

Figures (9)

• Figure 1: Block diagram of the hybrid semantic communication network scheme with separated learnable (top) and memorizable (bottom) data streams. The learnable data path is dedicated to the transmission of structured data $\mathbf{x}_l$ which will carry semantic information, while the memorizable data path is dedicated to the transmission of unstructured data $\mathbf{x}_m$ that is going to complement the semantic one.
• Figure 2: Shannon's communication paradigm in (a) compared to identification paradigm in (b), as illustrated in cabrera20216g. In (a) the receiver needs to the decode the received distorted code-word $c_i$ and retrieve an estimate of the transmitted message $\hat{x}_i$. In (b) the receiver needs to determine if the received distorted codewords set $C_i$ encodes a message that is relevant to him or not.
• Figure 3: Error types of message identification as in cabrera20216g. Type I errors represent the syntactic communication errors, leading to strong incorrect identification. Type II errors are created by the ambiguity present in the message identification paradigm and can lead to incorrect identification.
• Figure 4: Learnable data gets separated from memorizable data and mapped into an identity vector. Different shades of green represent different feature's value between identities.
• Figure 5: Mapping learnable data points $\mathbf{x}_{l,i}$ to their corresponding identities $\mathbf{i}_i$ and then in to semantic elements $\mathbf{s}_k$.