Research

In this page you can find information about my:

• current research in the context of the AKT and CROSI projects;
• past research in the context of my doctorate research on Deploying Ontologies in Software Design;

AKT and CROSI related research

Knowledge management is the focus of the AKT (Advanced Knowledge Technologies) project, a £8 million 6 year Interdisciplinary Research Collaboration (IRC) funded by the UK government. Southampton is a major contributor to AKT.

My current research focusses on semantic interoperability and integration. In particular, I am researching ontology mapping and merging on the Semantic Web. This is the main theme of the industrially supported project I led, CROSI. One of the emerging needs of the ambitious endeavour Semantic Web, is to resolve semantic heterogeneity. This old and unsolved problem becomes more apparent in the emerging Semantic Web where reasoning could involve externally defined data that might be described in semantically different ways. Resolving semantic heterogeneity is a key enabler for achieving interoperability. One way of enabling interoperability is to perform some sort of ontology mapping, where in its simplest form will be to map concepts between two ontologies. For example, in the context of the CROSI project, we developed a mapping system that acts as an aggregator of several matching algorithms. We deployed this system with success in various contexts. For instance, in the annual ontology alignment competition (OAEI'05) we were ranked among the top two systems for aligning correctly real world ontologies from the medical domain. Our key findings in this line of work are: (a) a Semantic Intensity Spectrum, which allows us to classify capabilities of ontology mapping systems and algorithms, (b) a modular architecture, which makes it possible to add and remove alignment tools in a seamless fashion, and (c) the CROSI Mapping System (CMS), which is our structure-based ontology mapping system. CMS is distributed as an open source software system and can be accessed by clicking here.

In my past work on this topic, together with my colleague Marco Schorlemmer (previously in AKT @Edinburgh, now at IIIA in Barcelona), we elaborated on a theory and developed a method for ontology mapping based on channel theory. We used the mathematical notions of classifications and infomorphisms to reason about the way local communities classify instances in their ontologies. Our aim was to capture the information flow between two ontologies, and the method we developed relies on semantics and classification of local instances rather than syntactic features as in most related work. We then used this information to automatically map them to global ontologies, these are top/upper level ontologies that each of the local ontologies conforms to and usually extends. In AKT, we have a similar ontology, AKT Reference ontology, which models a typical CS department in UK Academia. The whole idea is better illustrated in our Information Flow based Mapping system, IF-Map.

I am also interested in finding ways to design and build ontologies in (semi-) automatic ways after a domain analysis has been done. This is not the same as using documents and natural language technology to parse a corpus of document and extract some sort of structure which could be used in ontology building. My interest is focussed on more formal approaches for bottom-up construction of ontologies which could be synergetically work together with top-down approaches. Typically, the former are provided by domain users and/or data and the latter from domain experts and/or knowledge engineers. The good news is that there are formal approaches to assist developers for both: Formal Concept Analysis (FCA) is a data analysis toolset capable of providing the basic hierarchical structure of a domain and Description Logics have gained much respect and popularity in recent years as the right vehicle to use for modelling ontologies, especially those that are to be used on the Semantic Web. I am looking closely at the intersection of these two and exploring possible synergies.

In Southampton we have been working at the forefront of Semantic Web technologies. Recently, we launched an online RDF store containing information about the UK computer science community. The store, called AKT Triplestore, can be accessed here. There are a number of services which the triplestore provides, one of them, the award-winning CS AKTive Space. provides a knowledge navigation application for the UK CS community. Click here to fire a query which retrieves my personal information as it appears in the triplestore.

I am also involved in a technique we developed in Southampton for Ontology Network Analysis. We view the ontology as a network, and we traverse it with a spreading activation algorithm which is tuned to find the most important nodes. We measure importance in terms of popularity: the most well connected nodes are the most important. This information, is then used to help us identify Communities of Practice in the Academic domain. We recently applied this approach to automatically set up the content of an organisational memory, thus tackling the "cold start" syndrome faced when initiating an organisational memory.

In the first year of AKT research, I was in KMi, another AKT collaborating institution. I co-designed, developed and deployed the MyPlanet system. MyPlanet is an ontology-driven Web-based newsletter. Users submit an email story (e-Story) about a potentially interesting theme in the context of KMi lab. These are then annotated with knowledge structures drawn from the underlying ontology, rendered to produce a high quality Web page to display them, and users who might be interested in this e-Story are alerted via email. MyPlanet is the designated front-end which allows users to browse the e-Stories archive by specifying their preferences with respect to e-Story's contents. We categorized these contents in accordance to ontology-drawn knowledge structures normally found in an e-Story: research areas, themes, projects, technologies, application domains, organisations and people. We then store users' preferences as profile files, thus providing a personalised view of the knowledge containers' archive.

back to top

Doctorate research

From September of 1997 to June of 2000, I conducted my doctorate research at the Department of Artificial Intelligence in Edinburgh University.

Here's a brief description of my PhD: "Deploying Ontologies in Software Design":

My PhD, which was funded by the European Commission, aimed to discover ways of detecting errors in software design as early as possible. This, ideally, should be prior to the enactment of preliminary coding phase where any conceptual error conveyed by previous phases might be "absorbed" as a valid specification fragment, leading to pernicious side effects for the remainder of the software life cycle. To realise this sort of conceptual error checking we need to have available descriptions of domain knowledge to apply them to the model description. Ontologies are standardized descriptions of domain -consensual- knowledge, whereas the model descriptions are often viewed as executable, formal, specifications. The latter, when they are written in a declarative style(i.e.: using logic programming languages), make it possible to dispel ambiguities. Further, by being executable they represent not only a conceptual, but also a behavioural model of the system to be implemented. In addition, the declarative representation of ontologies makes it easier to apply them directly to this sort of specifications, thus "bringing domain knowledge in the application". 

An abstract of my thesis is available here. The whole thesis (207 pages) is available in zipped postscript format(638KB) or in unzipped postscript format(3.4MB).

My thesis supervisors were Dr Dave Robertson and Professor Austin Tate. 

Here are some key publications highlighting the contributions of my thesis:

"Use of Formal Ontologies to Support Error Checking in Specifications",
Y.Kalfoglou, D.Robertson
EKAW - 1999
more details...

"Managing ontological constraints",
Y.Kalfoglou, D.Robertson
IJCAI workshop on Ontologies and Problem-Solving Methods - 1999
more details...

"Using meta-knowledge at the application level",
Y.Kalfoglou, D.Robertson, A.Tate
DAI Reseach Paper - 1999
more details...

"Meta-knowledge in systems desing: panacea ...or undelivered promise?",
Y.Kalfoglou, T.Menzies, K-D.Althoff, E.Motta
The Knowledge Engineering Review - 2000
more details...

"Exploring ontologies",
Y.Kalfoglou
The Handbook of Software Engineering and Knowledge Engineering - 2001
more details...

"Maintaining ontologies with organisational memories",
Y.Kalfoglou
Knowledge Management and Organizational Memories- 2002
more details...

In my PhD days, I was a member of the Software Systems and Processes group

back to top

EU funding

My PhD project was supported by the European Commission under a Marie Curie Fellowship, Training and Mobility of Researchers(TMR), 36 months Grant.

back to top

SSP

In Edinburgh, I was a member of the Software Systems and Processes research group. Here's a link to group's research themes and publications. The group leader is Dr Dave Robertson. While in that group, I produced a catalogue for ontologies research which was presented at a panel discussion in SEKE'99. Together with my former colleague Daniela Carbogim, we presented an intricate issue to our SSP colleagues: "Challenges of the new E-ra: the Pandora's box for AI"? The material presented at these talks is available here.

back to top