Zhaoming Qiu

Towards a complete OWL ontology benchmark

Aiming to build a complete benchmark for better evaluation of existing ontology systems, we extend the well-known Lehigh University Benchmark in terms of inference and scalability testing. The extended benchmark, named University Ontology Benchmark (UOBM), includes both OWL Lite and OWL DL ontologies covering a complete set of OWL Lite and DL constructs, respectively. We also add necessary properties to construct effective instance links and improve instance generation methods to make the scalability testing more convincing. Several well-known ontology systems are evaluated on the extended benchmark and detailed discussions on both existing ontology systems and future benchmark development are presented.

EIAW: towards a business-friendly data warehouse using semantic web technologies

Data warehouse is now widely used in business analysis and decision making processes. To adapt the rapidly changing business environment, we develop a tool to make data warehouses more business-friendly by using Semantic Web technologies. The main idea is to make business semantics explicit by uniformly representing the business metadata (i.e. conceptual enterprise data model and multidimensional model) with an extended OWL language. Then a mapping from the business metadata to the schema of the data warehouse is built. When an analysis request is raised, a customized data mart with data populated from the data warehouse can be automatically generated with the help of this built-in knowledge. This tool, called Enterprise Information Asset Workbench (EIAW), is deployed at the Taikang Life Insurance Company, one of the top five insurance companies of China. User feedback shows that OWL provides an excellent basis for the representation of business semantics in data warehouse, but many necessary extensions are also needed in the real application. The user also deemed this tool very helpful because of its flexibility and speeding up data mart deployment in face of business changes.

Model-driven ontology engineering

W3C’s Semantic Web provides a common framework that allows data to be shared and reused across application and enterprise. As the Semantic Web shapes the future of the Web, it becomes more and more important in software engineering and enterprise application development. While existing ontology engineering tools provide a stack of ontology management support and are used successfully in certain domains, there still remains a gap between the ontology engineering tools and the traditional software engineering. For several decades, software engineering has been established on different modeling languages and methodologies such as Unified Modeling Language (UML). The differences in modeling languages and methodologies cause difficulties in enterprise application development involving the Semantic Web technologies. The existing ontology engineering tools provide only an ad hoc approach to bridging this gap with limited functionality and performance. The primary objective of our work is to bridge this gap between two different, but complementary engineering disciplines with a systematic approach. Our approach leverages Model-Driven Architecture (MDA) and Ontology Definition Metamodel (ODM), which enable model transformation. This approach allows seamlessly supporting existing models in UML and other languages in Semantic Web-based software development. In addition, it allows exploiting the availability and features of UML tools for creation of vocabularies and ontologies. Furthermore, MDA enables code generation and facilitates software tool development. This paper presents an MDA-based system for ontology engineering. In addition, it presents the entire stack of individual components of the developed ontology engineering tool.

Enhancing the open-domain classification of named entity using linked open data

Many applications make use of named entity classification. Machine learning is the preferred technique adopted for many named entity classification methods where the choice of features is critical to final performance. Existing approaches explore only the features derived from the characteristic of the named entity itself or its linguistic context. With the development of the Semantic Web, a large number of data sources are published and connected across the Web as Linked Open Data (LOD). LOD provides rich a priori knowledge about entity type information, knowledge that can be a valuable asset when used in connection with named entity classification. In this paper, we explore the use of LOD to enhance named entity classification. Our method extracts information from LOD and builds a type knowledge base which is used to score a (named entity string, type) pair. This score is then injected as one or more features into the existing classifier in order to improve its performance. We conducted a thorough experimental study and report the results, which confirm the effectiveness of our proposed method.