Announcing a holistic and comprehensive paper on semantic communications

We are delighted to announce a new holistic and comprehensive paper on semantic communications (SC) that Christina Chaccour, Dr. Saad, and collaborators have been working on for a little over a year now:

SC has attracted a lot of attention in our community recently, with the attitude towards it ranging from snubbing it as an "old" topic of no value (e.g., semantic communication is nothing but application-aware networking that we solved in the 2000s) to repurposing known concepts (e.g., compression or simple image feature extraction) as semantics all the way to describing semantic communication as the solution to all of the universe's problems for the next 25 years.

To address and reconcile this broad spectrum of perspectives, we have just written perhaps the most comprehensive paper on semantic communications that rigorously lays the technical foundations of the topic, by concretely explaining, with technical depth rather than qualitative discussions, how to build from an end-to-end perspective the next-generation semantic communication systems based on concepts from reasoning, information theory, transfer learning, minimum description length theory, and AI. This paper clearly answers several fundamental questions, including:

- What does semantic communication really mean (and what is NOT semantic communication)? What are the challenges? (See Sections I and II)

- How do we leap from today's data-driven networks, towards knowledge-driven semantic networks? (See Sections III and IV)

- What are the technical pillars of semantic communications? What are the advantages? (See Sections I and II)

- What does reasoning (a term that has been used and abused in the wireless literature so far) technically mean, what is its role for semantic communications, and what are the concrete AI tools that we can use to perform reasoning in semantic networks. (See Section V)

- How do we revisit key performance indicators, and more broadly, tools like information theory, to create semantic-based KPIs ? (See Section VI)

- How do we scale from semantic communication (link-level) to semantic networks, including integration with emerging platforms like ORAN? (See Section VII)

We are hoping that this paper will stir up the discussions, and also provide concrete technical foundations that allow semantic communications to truly become a revolutionary paradigm (for 6G and beyond), and not just another hyped up topic that will disappear in a few months (or be reduced to some form of compression or snubbed as “application-aware” networking).

Christina and Dr. Saad presented some of the material in a tutorial on SC at the IEEE 2022 GLOBECOM Conference on December 5th, 2022.

Posted on Dec14, 2022