The basic idea of the Semantic Web is to describe the meaning of Web data in a way suitable for automatic reasoning. Expectedly, the Semantic Web technology will bring about large-scale heterogeneous Web knowledge bases with a qualitatively new level of service. The concept of ontology (domain theory) will play a key role as a formal, explicit specification of shared conceptualizations that describe the semantics of data on the Web. Formal ontology languages as well as meta-level representation of Web resources are investigated. The possibility of developing automated reasoning systems for Semantic Web is explored from both theoretical and practical viewpoints, e.g., a hybrid approach with a strict separation between ontology predicates and rule predicates and a homogeneous approach embedding rules and ontologies in a logical language. Realization of the Semantic Web vision demands further research works on several other knowledge-representation-related issues.
Due to language-structure differences, when an information extraction (IE) framework that works well in one language is applied in a different language the framework often needs modification and supplementary components are often necessary. By incorporation of ontology-based semantic annotations and appropriate extraction filtering techniques, it is expected that IE based on patterns of triggering class tags and triggering plain words can be realized for Thai documents. We aim to develop a system for extracting semantic information from Thai text and representing extractions in a form of machine-processable frames. Extracted frames along with domain ontologies constitute a knowledge base from which logic-based reasoning engines can deduce several kinds of implicit information and relations. Application of this framework includes extraction of semantic frames from free-text abstracts of technical articles, which enables semantic document indexing and precise information retrieval based on semantic contents.
Reasoning with UML Diagrams
The Unified Modeling Language (UML) is a graphical language, adopted as a standard by the Object Management Group (OMG), for visualizing, specifying, constructing, and documenting the artifacts of a software-intensive system. As reported by recent works on the formal semantics of UML, there exist inherent interrelationships between components of a UML model. Such interrelationships constitute part of general knowledge about the domain of UML, which may be used for, e.g., deriving implicit properties and verifying the consistency of the model. A framework for knowledge representation and reasoning in the domain of UML is proposed, in which a UML model is represented as textual XML data, and the general knowledge about the UML domain as an XML declarative description. Development of an inference engine for automatic refinement of the encoded UML diagrams and derivation of implicit model properties is underway.
Equivalent-Transformation Computation Model
In declarative paradigms, a declarative description plays the role of a precise specification, and, at the same time, operates as a program. A number of works on amalgamation and generalization of declarative languages have been proposed. Most of them have been driven mainly by computation-oriented requirements, e.g., enhancement of operational semantics and integration of computation models; other important related concepts such as program synthesis and program transformation are investigated only afterwards and not inherent in their designs. By contrast, the equivalent transformation (ET) paradigm takes a program-synthesis-oriented approach, i.e., effective generation of efficient and correct programs from specifications is its underlying design motivation. Investigation of the ET computation model is in progress from both theoretical and application viewpoints .