Name:
|
Computational Approach to Ontology Profiling of Scientific Research Organisations
|
Description:
|
The profile of any scientific research organisation should include the major subjects on which the organisation works. Conventionally, the profiling is done based on a summarising statement developed by a committee, which may be subject to errors of human judgement. In this project, we propose to address the issue by an automated analysis of the individual members’ research outcomes and of matching them to ontology of the field of science in which the collective is engaged. In this way, a systematic and objective summary based on data integration, rather than on the aforesaid statements, can be achieved for both individual organisations and for combined regional science strengths. To our knowledge, no computational tool of such scope has been reported in the literature.
To match the structure of the scientific subjects, being pursued in the organisation with the conceptual structure of the scientific field, as reflected by its ontology, we first analyse the structure of interrelation between individual research subject items within the ontology relation, according to the organisation’s activities. This structuring will be achieved using both manual and automatic assessment of the publications by organisation members. For that, text mining techniques will be developed. Furthermore, a similarity measure between scientific items will be defined, on the basis of their place in the ontology, and the organization will be represented as a set of possibly overlapping subject clusters over the ontology. To exact such a representation, one important theoretical instrument of data analysis, the additive clustering model, will be adopted for both overlapping clusters and fuzzy cluster structures. These two approaches differently address the issue of overlapping attributes, one relating to individual belongingness (fuzzy clustering) the other to co-existence of many groups (crisp clusters overlapping). Developing of these clustering approaches will be based on the paradigm of data recovery, to allow for exploring the aspects of evaluation and interpretation of clusters.
Secondly, the so derived subject clusters will be parsimoniously mapped to an ontology relation of the corresponding field of science, and scientific
profile structures will be identified over the ontology, with the clusters’ head subjects as well as the missing, i.e. the absent scientific subjects within the cluster with respect to the ontology.
Finally, a set of open hypotheses will be explored regarding the interpretation and evaluation of organisational profiles, as well as regarding their aggregation at the regional or national level.
Thus, the project’s outcome will comprise a method, automated tools, and experimentally tested techniques for ontology profiling some scientific organisation, including the following:
(1) Identifying a set of ontology subject items and measuring similarities between them according to scientific activities of the organisation under consideration;
(2) Clustering the subject items into related groups;
(3) Mapping the groups to the field’s ontology, in order to derive the organisation’s profile;
(4) Aggregating and interpreting profiles.
The proposed methodology will be applied to scientific research organisations of computer science, specifically, Computer Science (CS) departments
of Universities in Portugal and UK, having as ontology of reference the ACM Classification System.
Detalhes COPSRO
|