Semantic Communication: From Philosophical Conceptions Towards a Mathematical Framework

2025-05-02

AI Summary

Semantic ambiguity induced by physical channel noise poses a fundamental challenge in semantic communication. Method: This paper establishes the first probabilistic semantic communication model grounded in a philosophical information perspective—incorporating constraint-based affordance and abstraction hierarchies—formalizing semantic content as domain-agnostic abstract constraints and orthogonally decomposing semantic transmission tasks. Contribution/Results: It rigorously unifies semantic communication with Shannon theory, proving the latter as a special case. A lower bound on semantic capacity is derived, demonstrating that the reliable semantic transmission rate is always at least the Shannon capacity, with an additive gain term equal to the conditional entropy (H(X|S)). This bound reduces to the classical Shannon capacity when no semantic ambiguity exists. The framework provides a theoretical foundation for modeling semantic-layer reliability and designing communication systems that transcend bit-level performance limits.

Technology Category

• Machine Learning: Information Theory
• Reasoning under Uncertainty: Relational Probabilistic Models
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Large language models for search

Abstract

Semantic communication has emerged as a promising paradigm to address the challenges of next-generation communication networks. While some progress has been made in its conceptualization, fundamental questions remain unresolved. In this paper, we propose a probabilistic model for semantic communication that, unlike prior works primarily rooted in intuitions from human language, is grounded in a rigorous philosophical conception of information and its relationship with data as Constraining Affordances, mediated by Levels of Abstraction (LoA). This foundation not only enables the modeling of linguistic semantic communication but also provides a domain-independent definition of semantic content, extending its applicability beyond linguistic contexts. As the semantic communication problem involves a complex interplay of various factors, making it difficult to tackle in its entirety, we propose to orthogonalize it by classifying it into simpler sub-problems and approach the general problem step by step. Notably, we show that Shannon's framework constitutes a special case of semantic communication, in which each message conveys a single, unambiguous meaning. Consequently, the capacity in Shannon's model-defined as the maximum rate of reliably transmissible messages-coincides with the semantic capacity under this constrained scenario. In this paper, we specifically focus on the sub-problem where semantic ambiguity arises solely from physical channel noise and derive a lower bound for its semantic capacity, which reduces to Shannon's capacity in the corresponding special case. We also demonstrate that the achievable rate of all transmissible messages for reliable semantic communication, exceeds Shannon's capacity by the added term H(X|S).

Problem

• Developing a probabilistic model for semantic communication grounded in philosophy
• Orthogonalizing semantic communication into simpler sub-problems for analysis
• Deriving a lower bound for semantic capacity considering channel noise

Innovation

• Probabilistic model based on Constraining Affordances
• Orthogonalization into simpler domain-independent sub-problems
• Derived semantic capacity lower bound exceeding Shannon's