Ontology Outreach Advisory

The second OOA Workshop at the Human Capital Summit in Maastricht has been a great success. Nearly thirty attendees, some from academia and most from industry, followed the programme which was compiled of two sessions and a wrap-up.

The first session contained presentations of some relevant competency-oriented models and frameworks. The second session saw a managed, lively discussion centered around a 'linking ontology' which could explicitly show semantic overlaps and differences between the various competency models/frameworks. The third wrap-up session was brief and positioned the OOA workshop inside the HR-XML domain.

First Session: Presentations

Robert Meersman opened the session with a brief introduction to the purpose and goals (.ppt) of the Ontology Outreach Advisory. Founded as one of the results of the KnowledgeWeb Network of Excellence, the OOA is destined to be a prime technology transfer vehicle from semantic research to industry.

The first year of the OOA already saw impressive achievements:

• The kickoff workshop in Oxford
• Scientific workshop OnToContent in Montpellier
• The publication of the first OOA-HR Roadmap (semantics in Human Resources)
• A number of published real-world success stories for dissemination
• The second OOA workshop in Maastricht (this workshop)
• The second OnToContent in Vilamoura, this November in Portugal
• Collaborations with other institutions in the Healthcare domain

Jeroen Hoppenbrouwers gave a quick introduction to the challenges of semantics and the difference between semantics and data models, or even information models.

The annotation of existing (legacy) systems, legal documentation, system requirements, and other artifacts including ontologies and models to shared knowledge structures is an area in which still much work needs to be performed. The shared knowledge structures can be ontologies, but in order to serve their purpose, they need to be both formal enough to be machine-processable and lightweight enough to be human-maintainable. This is a challenge for the next decade and essential for any further development of semantic applications, including the semantic web.

Luk Vervenne outlined the current developments surrounding HR-XML and its movement towards a semantically annotated framework for HR-related information exchange.

It has become clear that in order to reach any practical level of interoperability between (production, legacy) HR-oriented systems, a refocus on semantics instead of 'just syntax' is important. The introduction of explicit 'semantic linking pins' between existing HR-XML structures needs involvement from many stakeholders and is a combination of technical and organizational efforts. Some infrastructural work can be put into place rather soon, while the construction of linkup structures will need to be performed in lockstep with actual efforts to link up existing systems.

Andreas Schmidt | download presentation (.pdf)
Competency-oriented approaches are gaining ground in human resource development. Key technology to cope with the complexity of these approaches are ontologies, both for defining competency frameworks and concrete competency catalogs. Ontologies enable the automation of competency-related tasks like competency gap analysis, suggestion of learning opportunities or similarity-based profile matching for team staffing or job applicant selection. To provide a unified view on these issues at the crossroads of Human Resources Management and Development, knowledge management and informal learning and e-learning, we have developed the Professional Learning Ontology as a reference ontology, which is freely available. It provides reusable upper-level concepts that can be refined for concrete application cases.

For sustainability in companies, however, we need to go beyond mere technology and consider usage and update processes embedded into interwoven corporate processes. For that purpose, we have developed a reference process (providing continuous update processes connecting operational and strategic levels) which closely interlinks competence management and human resource development processes. This references process uses the ontology as a mediating artefact.

Tobias Ley | download presentation (.pdf)
Tobias' talk introduced a formal model of how to describe learning goals and prerequisite knowledge in a work-integrated learning environment. The model is based on knowledge space theory (KST) which has was originally developed in cognitive psychology. A competency-based extension of KST and the current status of application in the APOSDLE project was presented.

APOSDLE uses a data model which stores connections between tasks and competencies, as well as prerequisite relationships between competencies. The model is used to adaptively devise learning interventions which are integrated into the usual work tasks. A first implementation has been done in the domain of requirements engineering. The talk closed with some considerations for standardisation issues.

Clementina Marinoni | download presentation (.pps)
Why are competence and qualifications frameworks becoming more and more important within the Human Resources (HR) domain? In general, frameworks are necessary to achieve standardisation. In this case, frameworks help build a common language, i.e they help understand and communicate the same concepts.

Common language and understanding are a necessary premise to make services interoperable, i.e. each other connected, consistent, easily reachable, comparable and usable; interoperability means effectiveness and transparency and concerning HR domain, it fosters students and workers mobility, too. In order to develop common patterns it was necessary to define and to find an agreement on:

• The meaning of “competence” and “learning outcome”: i.e. what they are;
• Competence and learning outcome descriptions: i.e. how they can be described;
• Competence and learning outcome levels in line with the EQF levels (and with the European eCompetence framework logic): i.e. where we can position ICT competences and learning outcomes along the EQF eight-level scale.

Download the full presentation paper

Second session: Interactive Workshop

After the presentations of the individual models, which were for the most part geared towards a specific domain, application, or user group, the full workshop audience gathered to create the first draft of a linking ontology. Such an ontology is not intended to carry all knowledge contained in the individual models, but to convey the elements in the models that are equivalent in meaning.

Jeroen Hoppenbrouwers acted as workshop host to coin the first topical sentence, where he on purpose challenged the audience by bringing up a concept that was not apparent in most of the presented models: 'human being' aka 'Person'. Immediately a lively discussion started which quickly led to strings of concepts (in the DOGMA method for semantical modeling, these are called 'lexons') appearing on the large projection screen. Aggelos Liapis, also from VUB STARLab, documented the discussion live and for all to see; the result of this is the image below (click to enlarge).

Fairly quickly the 'Person' concept was pushed to the side, as the audience focused squarely on 'Competency' as the main topic of interest. It took a while before several members of the audience discovered that even such as basic concept had significantly different semantics in the minds of different people. In the end, agreement was reached that a 'Competency' is an artificially made-up conception of somebody's ability to perform a certain task. A 'Competency' therefore is not the actual ability, even if in some cases this actual ability also is named 'Competency'. Some more of these 'proxied' concepts were discovered during the session.

Concern was voiced that such a relatively abstract discussion and resulting model was too far from 'work floor reality' to be of use in industry. The point of the model was not to replace any other model, but to show where the existing models touch, and this is something that is not directly useful in for example job application systems. However, when making such existing (even legacy) systems interoperable, it is absolutely essential that the developers and users of these systems are sure of truly shared concepts between their own system and its peer systems. A discussion of the type performed in Maastricht proved to quickly reveal essential differences between a common concept such as 'Competency', potentially saving costly mistakes had an attempt been made to link up two differently conceived systems this way.

After an hour of hands-on (re)modeling and discussion, several issues were raised that need attention, both in research and for the next workshops:

• It probably is more efficient to limit the discussion to (and invite people based on) only two peer models/applications/domains at the same time. This could avoid 'hot spots' in the audience while others are relatively quiet.
• Before starting the workshop discussion, a clear set of discussion rules should be made available, explained, and afterwards enforced. Although some rules were introduced and used during the Maastricht discussion, the group dynamics sometimes moved out of focus.
• The documentation tool (a standard mind map) probably can be better tuned to the particularities of such a group process.

In the end, the workshop audience agreed that insight into the problem had been gained and many of the attendees favoured sessions such as these in the future (even outside the OOA context) to gather semantical differences between established communities of interest. The importance of making these semantical differences apparent was unanimously ranked as 'extremely high'.

Wrap-up

A very brief statement of the results presented above was given at a shared wrap-up session about HR-XML in general. People were enthusiastic about the general approach and even the result obtained during the workshop, no matter its relatively small scale. The process of getting to agree on what commonly used concepts actually meant in the various competency-related models was considered helpful, practical though not yet fully mature, and certainly an approach worth further development.