The field of Semantic Computing was introduced by Phillip Sheu in 2007 with the launch of the IEEE International Conference on Semantic Computing (ICSC) [1] and the International Journal of Semantic Computing (IJSC). As articulated in the inaugural issue of IJSC [2], “the field of Semantic Computing addresses computing technologies, and their interactions, that can be used to extract or process the contents and semantics of active services and passive data that are unstructured, semi-structured, as well as structured.” It extends the Semantic Web—traditionally focused on ontology-based augmentation of web pages—both in breadth, by incorporating multimedia, services, and structured data beyond the web, and in depth, by addressing the access, use, synthesis, integration, and analysis of data and services. In doing so, Semantic Computing unifies diverse areas such as software engineering, user interfaces, natural language processing, artificial intelligence, programming languages, grid computing, and pervasive computing within a coherent framework.

​

This vision is realized through a multi-layered architectural model designed to process, integrate, and utilize semantic information. The original framework consists of four layers: Semantic Analysis, which interprets signals such as pixels and words to extract meaning; Semantic Integration, which consolidates content and semantics from heterogeneous sources; Applications, which leverage this semantic content to address specific tasks and may provide services to other applications; and the Semantic Interface, which enables users to access and manipulate semantic content across sources. This architecture was subsequently refined into a five-layer model [3], in which the Applications layer was divided into Semantic Services—focused on solving specific problems through capabilities such as web search, question answering, content-based multimedia retrieval, and semantic synthesis—and Service Integration, which orchestrates multiple semantic services to deliver more comprehensive and interoperable solutions.

​

The book Semantic Computing and AI [4] chronicles key milestones in the evolution of the field, highlighting foundational theories, emerging trends, and real-world applications across the closely related domains of semantic computing and artificial intelligence (AI). Drawing on a curated selection of articles from IJSC published between 2007 and 2025, the volume traces the growing convergence of these disciplines and their collective impact on advancing machine intelligence.

Rapid advances in computing have driven significant progress in both semantic computing and AI, enabling new approaches to data processing, knowledge representation, and intelligent decision-making. Their intersection has become increasingly important, offering powerful methods for addressing complex computational challenges and enhancing how machines interpret and interact with human knowledge. As such, this body of work provides a valuable resource for researchers, practitioners, and students seeking to understand both the evolution and the broader implications of these fields.

​

Recent developments in Semantic Computing include Generative Problem Solving (GPS) [5][6], a paradigm that supports not only the solution of computational problems but also the generation of new problems. Although both GPS and generative AI involve forms of “generation,” GPS extends the concept by creating problems, data, and use cases to facilitate systematic problem solving. In contrast, generative AI focuses primarily on producing novel outputs that reflect patterns learned from textual and multimodal data. GPS emphasizes a step-by-step approach in which problems are generated and solved iteratively, enhancing transparency and enabling users to follow the reasoning process. When combined with human-in-the-loop strategies, GPS has the potential to significantly improve both learning efficiency and solution quality.

​

Despite the remarkable natural language capabilities of large language models (LLMs), important challenges remain. In particular, LLMs often struggle with tasks that require rigorous logical reasoning or precise computational procedures. They may fail to systematically explore the full space of possible solutions or to guarantee correctness through explicit, step-by-step reasoning. GPS seeks to address these limitations by augmenting LLMs with structured, transparent, and verifiable problem-solving capabilities.

​

1. Message from General Chairs". International Conference on Semantic Computing (ICSC 2007). 2007. pp. xiv. doi:10.1109/ICSC.2007.4. ISBN 978-0-7695-2997-4.

2. Sheu, Phillip C.-Y. (2007). "Editorial preface". International Journal of Semantic Computing. 01 (1): 1 9. doi:10.1142/S1793351X07000068. ISSN 1793-351X.

3. Sheu, Phillio C.-Y.; Ramaoorthy, C. V. (2009). "Problems, Solutions, and Semantic Computing". International Journal of Semantic Computing. 03 (3): 383–394. doi:10.1142/s1793351x09000781. ISSN 1793-351X.

4. Florian Schimanke, Mustafa Sert, and Chung-Hsien (eds), Semantic Computing and AI, World Scientiic Publishing, 2025

5. Hsiang-Shun Shih, Chengheng Lyu, Goli Vaisi, Phillip C-Y Sheu, "Generative Problem Solving," Proceedings, International Conference on Semantic Computing (ICSC 2025) (PDF)

6. 20th anniversary special issue, International Journal of Semantic Computing, 2026 (to appear)