2013
As part of the ORE 2013 organisation, I spearheaded the key event of the workshop: the system competition. This involved designing and implementing a reasoner benchmark framework, which tests both performance and soundness/completeness of DL reasoning systems. Additionally, I carried out the entire benchmark, from preparing test machines to determining winners.
2012-13
During my PhD I did a short placement as a research associate, in an EPSRC project headed by Robert Stevens. I developed an ontology to describe scientific equipment, underlying technologies, functions, and locations. Then I implemented a prototype Web application [this one] that allows browsing equipment in a faceted way.
2011-13
This project aimed at identifying whether an ontology is performance-heterogenous for a given reasoner, that is, whether it contains performance-degrading sets of axioms. We invented techniques to identify such sets, so-called ontology hotspots, analogous to hotspots in software engineering; i.e. a fragment of code that is a performance bottleneck for a particular program (the program in our case being the reasoner). Our techniques show significant improvements over the original classification performance (at least 5x faster) when hotspots are approximated.
2010-13
Part of my research while in Manchester resulted in ecco: a hybrid diff tool for OWL ontologies that (in a nutshell) detects and aligns axiom changes between ontologies, and aligns these with concepts whose meaning they affect. The related paper Concept-Based Semantic Difference in Expressive Description Logics was runner-up for Best Student Paper Award in 2012, given out by the School of Computer Science. A Web-based front-end for ecco can be found here.
2010-13
Much of the work I have done uses the NCI Thesaurus as a test corpus; the mere number of publicly available versions makes it a valuable resource for ontology evolution studies. A challenging ontology (size and complexity-wise) as it is, the NCI Thesaurus is also a good candidate for reasoner benchmarks (of which I did a fair share). Hopefully I will port more (updated) results into a Web-friendly format, meanwhile, here's some corpus statistics and diff results.
2009-13
During my 1st year as a PhD student I developed a module extractor that guides users in choosing seed signatures for modules. It was appropriately called Hancock, Bijan's idea (all rights reserved). Later I developed another module extractor specifically dedicated to SNOMED CT and CORE Problem List Subset of SNOMED CT based signature files. Finally, I contributed to the Web-based module extractor.