Concept Mining for Data Management

The PhD project addresses data ambiguity, a core issue hampering the management of organisational data, and aims to support data management through concept mining. We intend to (semi-)automatically analyse perceived meanings of terms and concepts within an organisation’s data domain to aid terminological disambiguation. The concept mining system will deploy text mining and information retrieval techniques within unstructured data. We adopt a 3-phase approach of exploring the problem and requirements, iterative development and testing, and finally, an extrinsic evaluation in a real-life setting. Every component of the system will be intrinsically evaluated with metrics best suited to that component. The toolkit will also be evaluated qualitatively through studies with data managers for efficiency, quality and real-world applicability. The project will bridge the gap between domain expertise and automated methodologies, with expected contributions such as an evaluation framework for concept mining and advancement towards the research of terminology extraction.