Automatic Construction of Lightweight Domain Ontologies for Chemical Engineering Risk Management

Curtin University of Technology, Department of Chemical Engineering, GPO Box U1987,Perth, WA 6001, Australia; tel. +61-8-9266 7581, fax. +61-8-92662681, email:{N.Balliu,H.Wu,M.O.Tade}@curtin.edu.auThe need for domain ontologies in mission critical applications such as risk management andhazard identification is becoming more and more pressing. Mo st research on ontology learningconducted in the academia remains unrealistic for real-world applications. One of the mainproblems is the dependence on non-incremental, rare knowledge and textual resources, andmanually-crafted patterns and rules. This paper reports work in progress aiming to address suchundesirable dependencies during ontology construction. Initial experiments using a workingprototypeof the systemrevealedpromisingpotentialsin automaticallyconstructinghigh-qualitydomain ontologies using real-world texts.1. IntroductionHazard identification is a crucial aspect of risk management . The identification of hazards isthe prerequisite step to the analysis and treatment of risks. As such, clear definitions on the typeof risks and the processes involved for hazard avoidance and treatment are necessary. Unam-biguous definition enables effective communication, which is crucial in passing on experiencesand expertise to trainees and students dealing with dangerous chemicals and products. However,very often such knowledge is embedded in the domain experts’ mind, or scattered in variousformat, e.g. operation notes, online resources, scientific publications or technical reports. Anintegrated knowledge structure known as an ontology is therefore becoming necessary for de-scribing the concepts and processes to ease the process of information sharing and reuse. Somepossible applications of domain ontologies include conceptual document retrieval and decisionsupport system. The importance of ontologies to knowledge-based applications has promptedan increase in efforts to construct and maintain such knowledge structures. Generally, there aretwo ways of constructing ontologies, namely, manual crafting and automatic discovery.Manual construction and maintenance of ontology is often critised for being labour intensive,biased and static. Such manual process typically requires multiple domain experts to identifythe key concepts and processes, and then collaborate with knowledge engineers for effectivedigital representation. The neutrality and representativeness of manually-crafted ontologies isalso disputable when the domain experts are unable to reach consensus during the knowledgeengineering process. New changes to the domain are often ignored and cannot be incorporated