Semantic Communications via Features Identification
Federico Francesco Luigi Mariani, Michele Zhu, Maurizio Magarini · 2025
Read Semantic Communications via Features Identification by Federico Francesco Luigi Mariani, Michele Zhu, Maurizio Magarini (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-20720.txt.
Introduction
Federico Mariani, Michele Zhu, and Maurizio Magarini set out a new approach to semantic communication, in a paper from 2025. In this field, senders aim to pass meaning, not exact bits. The rise of sixth-generation wireless work has drawn fresh interest to the idea. Yet the field still lacks a clear, workable frame for how to do it.
Their method names a message by its key features rather than resending it whole. The slack and blur inside meaning can help pin down which message was meant. They test the approach with machine learning tools in simulations. Results show clear gains over classic bit-by-bit sending. The team frames the work as a bridge between two schools of post-Shannon thought. It gives engineers a concrete element to build semantic systems around.
Catalog record
| CREATOR | Federico Francesco Luigi Mariani, Michele Zhu, Maurizio Magarini |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 31 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-20720
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Semantic Communications via Features Identification Federico Francesco Luigi Mariani, Michele Zhu, Maurizio Magarini 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. I Introduction Soon after Shannon’s milestone publication [ 1 ] , Weaver [ 2 ] framed communication as a three-level problem: the technical transmission of symbols, the semantic problem, and the effectiveness problem. The first level, the technical aspect, has been extensively studied and applied in conventional communication systems, which are now approaching the capacity limit defined by Shannon due to the growing demand for a fully connected and integrated world. However, the ultimate scope of communication is often to transfer a specific meaning or intention [ 3 ] , regardless of the exact wording or symbols used to do so. This goal often clashes with the limitations of physical communication media —such as optical fibers, radio waves, or cables— that are optimized for the transmission of classical, syntactic , information. Semantic communications aim to overcome these limitations by focusing on transmitting only the semantically relevant information for a given task, which can significantly reduce data traffic in future wireless networks. Recently, the development of new communication technologies based on artificial intelligence (AI) has garnered growing interest from both industry and academia [ 4 ] , driven by advancements in key enabling technologies such machine learning (ML), large language models (LLM) [ 5 ] , and statistical modeling. These technologies are expected to play a central role in the upcoming sixth generation (6G) of wireless networks [ 6 ] , where semantic communications and intelligent networks have been identified as core challenges [ 7 ] . At the birth of information theory, the semantic aspect of communication was deliberately left out, focusing solely on the mathematical foundation of symbol transmission. Later works, including those by Bar-Hillel and Carnap [ 8 ] , sought to introduce preliminary definitions of semantic information within the context of language processing, yet a unified mathematical model that fully captures the nuances of semantic communication has remained elusive. Today, theoretical approaches to semantic communications are many and diverse, such as Ma et al. [ 9 ] defining universal semantic entropy and semantic channel capacity, Zhao et al. [ 10 ] [ 11 ] modeling the problem using simplicial complex structures, or Kolchinsky et al. who studied semantic information via a viability function [ 12 ] . Gaps between these different approaches need to be bridged to harness the full capability of semantic communications. Contributions: This work aims to consolidate and link the diverse approaches to semantic communications together, and also define a suitable semantic element for next-generation communications that can encapsulate all of the peculiarities of semantic communications and be reasonably practical in its use. Specifically, we investigate a combination of the Teacher-Apprentice communication framework [ 13 ] with the message identification paradigm [ 14 ] . Our goal is to propose a new communication method, named " Identification via semantic features ," which aims to exploit the ambiguity present in semantic communications to identify semantic elements by transmitting their features instead of the whole message. The proposed approach is tested through simulations and numerical analysis to demonstrate its potential for significant savings in transmitted bits, while still achieving high accuracy in identifying semantic elements. Organization: The paper is structured as follows. Section II overviews relevant works needed for our research. A detailed description of the proposed communication method is given in Sec. III . A test case with its numerical results is presented in Sec. IV followed by conclusions and further developments in Sec. V . II Semantic architecture and message identification The shift in the communication paradigm of semantic communications moves the focus from low-level bit accuracy to higher-level semantic fidelity, which introduces several issues and challenges. Ambiguity is one of the most relevant, especially in natural languages [ 15 ] , where a message can have different meanings depending on the context of communication. This creates significant challenges for decoding in semantic communication systems, making it difficult, if not impossible. Another major issue is the definition of the structure and atomic element that constitute a semantic message. In classical communication, the structure of a message is predefined in codewords, with the bit serving as its atomic unit. However, in semantic communications, there is no established agreement on the structure or basic unit that constitute a semantic message. In the following subsections, we will present the key literature that guided our work in addressing these issues, providing the building blocks for our research. Transmitter (Teacher) \mathbf{x} Semantic Encoder Encoder Noisy Channel Decoder Semantic Decoder Receiver (Apprentice) \mathbf{\hat{x}} Dynamically updated SeB Semantic Base Semantic Base Noisy Channel Encoder Decoder \mathbf{x}_{l} \mathbf{\hat{x}}_{l} \mathbf{x}_{m} \mathbf{\hat{x}}_{m} Fig. 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. II-A Semantic Communication Architecture The definition of a semantic communication framework helps us addressing the problems of ambiguity and the need to define a semantic element’s structure. Focusing on the physical layer, as illustrated by Zhang et al. [ 16 ] , a semantic communication framework includes several critical components that differentiate it from classical communication architectures. These include: the semantic base (SeB), which is responsible for establishing a common communication background between the transmitter and receiver, addressing part of the ambiguity problem; and the semantic encoder/decoder pair , responsible for encoding and decoding raw data inputs into semantic information. In semantic communications, the transmitter and receiver each possess a local knowledge base, which may differ at any given time. To address this, a third element — the dynamically updated SeB — must be introduced to synchronize the knowledge bases, allowing coherent semantic decoding at the receiver’s side. Building on these elements, Chaccour et al. [ 13 ] redefined the communication architecture by introducing the concept of a hybrid semantic-syntactic network .