Background: In order to improve ontology quality, tool- and languagerelated tutorials are not sufficient. Care must be taken to provide optimized curricula for teaching the representational language in the context of a semantically rich upper level ontology. The constraints provided by rigid top and upper level models assure that the ontologies built are not only logically consistent but also adequately represent the domain of discourse and align to explicitly outlined ontological principles. Finally such a curriculum must take into account the pre-existing skills and knowledge of the target audience. Objective: To develop a well-structured curriculum aligned to the particular requirements of life science professionals, in order to enable them to create logically sound, domain adequate and predicable ontologies using the Web Ontology Language (OWL) in Prote´ g e´. Methods: Content selection for the curriculum was based on the literature, pre-existing tutorials, and a guideline for good ontology development (i.e ontology design enhancing domain adequacy, sustainability and interoperability) that drew on the authors previous experiences with large ontology development projects. Learning objectives were formulated according to a needs assessment of the targeted learners, who were students trained in life sciences with basic knowledge and practical skills in computer science. As instructional format we choose an approach with a high amount of practical exercises. The curriculum was first implemented with 24 Students and 7 lecturers/ tutors over 5 full days. The curriculum was evaluated by gathering the participants feedback via a questionnaire. Results: Curricular development produced 16 modules of approximately 2 hours each, which covered basic principles of Applied Ontology, description logic syntax and semantics, as well as best design practices outlined in ontology design patterns and variants of the BioTop upper ontology. An opinion survey based on questionnaires indicated that the participants took advantage from the teaching strategies applied, as they indicated good knowledge gain and acknowledged the relevance of the modules. The difficulty was rated slightly lower. Conclusion: The development of teaching material for principled ontology design and best practices is of crucial importance in order to enhance the quality of biomedical ontologies. Here, we present a curriculum for a week long workshop, leveraging on current educational principles, focusing on interactive hands-on exercises, group interactions, and problem-oriented learning. Whereas evaluation clearly showed the success of this approach, in particular regarding student's satisfaction, the objective measurement of traceable effects on the quality of the generated ontology, although of much higher interest, has just started.
Ontology engineering continues to be an area of major interest
within the life sciences, as computer-interpretable domain
representations are the only viable option for an efficient and intelligent
exploitation of the vast amounts of high-throughput-generated data.
Although numerous ontologies are publically available through
ontology libraries and access portals, they remain of
heterogeneous quality and ontological rigor
While these efforts mainly target already experienced ontologists, only few efforts have recently focused on teaching basic notions of ontology engineering to novices. This imbalance results in a growing number of practitioners being forced to apply particular design requirements and patterns as requested by the aforementioned policy providers, but without even knowing the most basic foundations in the semantics of the representation language used.
Another major obstacle in ontology design is that various
scientific and engineering communities sustain different and sometimes
even contradictory modeling objectives and paradigms. Although
skilled logicians or computer scientists should normally understand
the formal semantics of description logics, their approaches to
certain modeling tasks are often based on other paradigms, like
database technologies or object-oriented programming. Object-oriented
programming uses, for example, inheritance principles that
differ from the inheritance principle of description logics. While the
inheritance principles of modern object oriented programming are
based on the substitution principle
In the light of the rising demand for of formal representation of,
e.g., biomedical knowledge
In order to arrive at a sustainable development of sound ontology artifacts, it is necessary that there are ontology developers with (1) a technical background in computer science that will (2) closely cooperate with experts for the domain to be represented, and have (3) a foundational grounding in the principles of Applied Ontology and the logical formalisms underlying ontology description languages. The basic idea of our curriculum is to teach such principles as a series of robust and well-calibrated ontological building blocks, targeting students in life sciences disciplines with some background in computer sciences. It should teach the essential skills to novices to develop sound OWL-DL ontologies based on approved methods in formal ontology.
Some educational material on formal ontology engineering for non-computer scientists is publicly available, and widely used for self-training or in tutorials, esp. the Pizza Ontology tutorial (Prote´ge´-Tutorial)1 based on the Pizza ontology2 and “Ontology Development 101”3 or publicly available course material4. These tutorials are well structured and found wider acceptance, but suffer from some limitations. They are often written as quick introductions and user guides to the Prote´ge´ OWL editor and only partially address more complex practically occurring modeling problems, but rather focus on easy-to-understand examples. They do not teach how the ontologies created should be structured, and which tools lend themselves for sustainable design decisions.
As the recent debate on the scientific value of formal ontology in
the life sciences has shown, the development of real life ontologies
applying decidable logics is an interdisciplinary and highly
complex process, whose inter-dependencies are not yet fully understood
Several challenges have to be met in the ontology engineering process, starting with the usage of editing and reasoning artifacts with all their idiosyncrasies and computational requirements, followed by putting into practice the collaborative editing of “single-file” or modularly structured ontologies.
The scientific community has made considerable progress in the understanding and development of large ontologies. Besides new efforts in continuing the generation of knowledge, we must understand how to educate students to become ontology engineering experts. It is not acceptable that just a small community is able to understand a topic of high interest, which, in contrast, is often considered rather esoteric by those who are supposed to use the artifacts created. In our view, well-instructed domain experts are needed to build good ontologies, because they are the only ones who know what has to be represented and how, and who will later engage in the dissemination and use of ontology-enabled tools. Domain experts with some computer science background should be educated appropriately to understand the basics of philosophical and logical foundations needed for ontology development and be trained in real-life ontology engineering. 2
The development of a specific curriculum for building OWL-DL ontologies is part of a larger project in which the key elements for the design of qualitatively good ontologies are defined (GoodOD: Good Ontology Design). In this regard, the curriculum objectives and contents mirror major parts of the goals of the GoodOD project.
Initiation to ontology design faces the problem that every community involved has its own terminology. There are the vocabularies of philosophical ontology, set theory, DL, Prote´ge´, knowledge management, and so on. This can easily lead to ambiguities and accordingly to confusion on the side of the learner. We decided to choose a rigid front end orientation and to use Prote´ge´ vocabulary wherever possible, as this is what modelers see when working with this editor.
The curricular development followed a widely adopted method from
medical education
Based on a general requirements analysis as given in the background section, we identified the targeted learners of the curriculum as employees or students in the life sciences with a background in computer science, either as a minor subject or part of a bioinformatics curriculum. The needs of these learners with regard to ontology development were assessed.
Prior to the curriculum development, a guideline had been developed, in which the authors elucidated the principles of good biomedical ontology design using a decidable description logics in OWL format. The objective of the guideline is to provide practical guidance for novices and experts on how to use the abovementioned representation framework, and how to address ontology engineering projects using top-level ontologies and ontology design patterns. Based on this guideline and in view of the curriculum time constraints the most appropriate content for the learning objectives was selected.
The main step in curricular development is the formulation of
goals for the complete curriculum and specific educational
objectives. Based on the problem analysis, and the general and targeted
learners needs assessment, learning objectives and goals for the
curriculum were specified. Although literature-based, the final
selection of educational objectives was led by personal experience in a
series of life science ontology development projects
We decided to use the Prote´ge´ editor because of its free availability and its support for OWL and automatic reasoners. In view of our guidelines we decided, however, not to use all available features of Prote´ge´. For example, we did not include individuals and data type properties, because they are of no relevance for the modeling of proper ontological facts and even apt to mislead a novice developer.
A small sample ontology (Zoo ontology) was developed as a
running example throughout most lectures and exercises. We decided
to use the BioTop domain top level ontology
As the most appropriate instructional format for the training of skills practical exercises were developed, so that the curriculum structure was designed to be based on 16 training modules (each lasting 2-3 hours) which consisted in a short introductory oral presentation of about 15 min followed by one or more practical tasks. Hands-on exercises were shaped for pair-wise execution, whereas paper-based practical exercises were adapted to groups of up to six students.
Curriculum evaluation was performed with a questionnaire consisting of 48 closed, and 12 open questions, where the first ones assessed how students judged the educational principles, the difficulty of the content, and the relevance of each module using a five-point Likert scale. Additionally opinions and attitudes with regard to problems and advantages of the course were collected.
The curriculum was implemented in Summer 2011 as an elective summer school at Freiburg University, Germany. 24 students from Austria, Germany and Slovenia participated who either studied biology as major subject and computer science/ bio-informatics as minor subject or studied computer science/ mathematics as major subject and biology as minor subject. The participants agreed in an informed consent to the scientifically analysis of results of the summer school and to their participation in a subsequent educational study. Each participant received an expense allowance of e 500 after completing the summer school and the study. It was clearly communicated that the payment of the expense allowance would depend sn Module 16 – Spatial disjointness ODP r Representing organ parts with the spatial disjointness ODP ltygo ttaeP Module 15R–epCrleosseunretinOgDaPn animal taxonomy and use the closure ODP o nO isgn Module 14 – Introduction in Ontology Design Patterns (ODP) eD Understang ODPs and using the exception ODP
Module 13 – Information objects l Representing plans and documents on animals l-vee iseModule 12R–epNroense-mntaintegrhiaalbpithaytssicaanldoebnjeclcotssures of animals p goModule 11 – Collective entities to lo Representing feed and groups for animals isgn tnoModule 10R–epPrreosceenstsinagndlopcoamrtioctiipoantiaonnd development of animals U Module 9 – Introduction in the BioTop domain top-level ontology
Using the basic features of BioTop
Module 8 – Typical ontology design errors yg Learn to ask the right questions in building an ontology lo Module 7 – Description Logic reasoning tno n Using a DL reasoner in the editing cycle in Protégé o igModule 6 – Introduction in OWL and the Manchester syntax la se Use the restrictions editor in Protégé itcc dModule 5 –UsRienlgatrieolantsioannsdamnderiemoplolegmyenting a partonomy raP Module 4 –ImDpislejominetnstianngddpisojolyinhtiesraanrdchpieoslyhierarchies with Protégé Module 3 – The ontology editor Protégé
Implementing an is_a hierarchy with Protégé s isc liepModule 2 – Classification and Taxonomy a cn Building an is_a hierarchy B i r pModule 1 – Introduction in ontology and philosophical background on their uninterrupted participation, but not on results in any of the associated tests and questionnaires.
The condensed main curriculum took place in the first five days of the summer school. In the remaining three days a quantitative study on ontology development was performed.
Figure 1 shows the structure of the curriculum and the main contents of the modules. Sixteen modules with a length between two and three hours followed each other with increasing complexity. These could be grouped in four sections without sharp borders: Basic principles, Practical ontology design, Using top-level ontologies , and Using ontology design patterns.
Especially in the first phase of the curriculum, many exercises were conducted in group work without a computer, the aim being to demonstrate that major parts of the development process consist of cognitive decision steps, which are independent of technical skills or software programs.
Most modules were organized in a similar structure, consisting of an introductory short presentation (at most 20 min), followed by a training phase in which the students worked on practical tasks alone or in pairs. In a plenary session results, suggested solutions and problems were presented and discussed, and a take-home message was formulated. All material and all exercises were laid out in a brief document, which focused on most important details without disturbing the students’ creativity. Module 10 Process and participation is presented here as an example.
The module starts with a presentation which introduces processes as an important ontological category, relating it to the corresponding BioTop classes and relations: Processes (Process) are things which have parts in time following each other in a sequence, so that all process parts which are preceded by the next part in sequence are existentially related to the latter (preceded-by ). Processes are only fully instantiated when they are finished. They have at least one participant (expressed by the relation pair hasParticipant, participatesIn), and they have a duration (hasDuration). The location (expressed by the relation pair hasLocus, locusOf) of a process has to be differentiated from its participant. The process participation can be further distinguished, which is expressed by the subrelations hasAgent, hasPatient, and hasOutcome. A plan (Plan) to do something can only be realized (hasRealization) by processes.
In the four exercises of this module, practical skills and implementation issues had been theoretically introduced beforehand, and were then issued as practical Prote´ge´ editing task. To provide students with a framework of basic classes and relations, they were requested to use a custom-tailored BioTopLite version, that included basic constraints during practical ontology building tasks, and limiting potential variants to facilitate the subsequent comparison of the resulting OWL artifacts.
In the first exercise of this module students had to provide a taxonomy of animal locomotion: Swimming, Flying, Running, Riding and Digging. These processes should be represented with at least one participant and should be correctly localized in a suitable environment, e.g. (PortionOfAir, PortionOfLand, PortionOfWater). The definition of locomotive processes should then be used to define FlyingAnimal and the other classes described along their locomotion modality. In other exercises of this module the sequence of processual parts in a developmental process had to be represented with the precededBy relation or different roles had to be assigned to participants in a hunting action (Hunting, Hunter, Prey), so that individuals that are members of the same class can have different roles.
The quantitative evaluation of students on the curriculum was performed on a 5-point Likert scale (1 = very good; 5 = very poor).
As shown in table 1, students evaluated the overall didactic principles of the curriculum modules and their relevance positively. The ratings for difficulty were slightly lower.
As anticipated, the qualitative answers of the students showed a large variety of partly contradictory opinions and feelings on the curriculum. However, they purport a clear view on the following aspects of the curriculum:
The curriculum started too slowly regarding the amount and the difficulty of the module content. In the mid-curriculum the pace was evaluated to be just right, but accelerated too steeply at the end, where the more difficult issues had been addressed
in little time and more transfer tasks were demanded from the participants.
Even after eight days of hand-on work with Prote´ge´ , many students expressed lack of confidence in applying ome of its functionalities appropriately.
Many of the students were highly motivated by the curriculum and expressed interest allocating further time to ontology engineering.
Some students wanted to be trained more thoroughly in formal and logical backgrounds. Many students complained of “philosophical deviations” during presentations and wanted to have a more straight forward teaching focusing on clearer statements on what had to be done practically.
In the opinion of a few students the practical usage of ontologies in different scenarios should have been shown (and even practiced). For them the benefit, meaning and technical structure of an “ontology driven semantic framework” remained unclear and should have been incorporated in the curriculum. 4
This work is based on the assumption that better educational programs for the training of life sciences ontology engineering are necessary to improve the quality of the ontological artifacts produced in this domain. Consequently, we developed a curriculum which covered the most important aspects of ontologies in the biomedical domain according to the literature, based on prior work on a ontology design guideline and our practical experience as ontology developers. Moreover, the curriculum was specifically targeted to learners with a biomedical background, additionally trained in basic and practical computer science. This profile characterizes in our view the most important stakeholders in biomedical ontology building, maintenance, deployment, implementation, and use. The curriculum has consequently been developed to serve as a hands-on guide to real-life ontology design problems, and supports the future ontology engineer throughout the ontology development life cycle.
Although students were generally very satisfied with the curriculum (as shown in the evaluation), they also signaled some major and minor curriculum improvements. If their assessment on time allocation (in the beginning too long, in the end too short) is correlated with the curriculum structure (see Fig. 1) this could mean that the introductory modules, especially on taxonomy, should be condensed. On the other hand, the modules on complex real life scenarios, beginning in the middle of the curriculum, should have been allocated more time. Some students also wanted to gain deeper insight in formal and logical aspects, which were deliberately presented superficially, as we did not consider them important for a practical approach under the given time constraints. In the future this could be addressed by providing additional readings for those interested in these topics. As a consequence of the survey, we will modify the curriculum by integrating short tutorial sections about the principle and basic subject matter into modules with more practical impact. In addition to shifting the focus more towards practical design problems, we consider developing a new module which addresses the logical and formal foundations of ontology and description logic.
The modularized approach of the curriculum gives the educator freedom for customization, aligned to either specific requirements of certain user groups or to curriculum length. It is relatively easy to focus the modules on certain topics or to compress them for students with more pre-existing skills and knowledge.
There are two major sources for further development and improvement of the curriculum. Students evaluations as described above and typical mistakes and errors made in the exercises during the summer school. The latter were documented in detail and are currently subject of a systematic analysis. Future work will also include the design of online modules on the basis of the current curriculum, which enables the design of flexible courses in a blended learning approach, i.e. as a combination of electronic with face-to-face teaching. 5
The dissemination of good ontological practice in the life sciences is not only a matter of research, but also of the availability of professional trainers who impart the knowledge in this highly interdisciplinary area. On this basis, a curriculum of 16 modules was developed to train students in the biomedical domain with a background in computer sciences how to practically develop and build ontologies in OWL using the Prote´ge´ editor. The implementation of this curriculum in a summer school setting including 24 students clearly showed that it is adequate to convey and train the complex knowledge and skills of ontology engineering in a week long course.
This curriculum represents an outline, how to successfully train students and researchers in the life sciences with the abstract matters of ontology engineering, enabling them to formally represent their domain knowledge with maximum profit.
This work was supported by the Deutsche Forschungsgemeinschaft (DFG) grant JA 1904/2-1, SCHU 2515/1-1 within the project GoodOD (Good Ontology Design), http://www.iph.unirostock.de/Good-Ontology-Design.902.0.html.