Faculty of Computer Science

Research Group Theoretical Computer Science

Oberseminar: Heterogene formale Methoden

Abstract:

Access to huge data is required for an appropriate structure and grouping of data so that the access to the data becomes easier, the status which clustering algorithms usually accomplish. Special attentions are paid in recent years on semantic data clustering which is needed for semantic interpretation of the input data. Here, three modified clustering methods are used and the results of these techniques are evaluated. Based on this, first a technique is developed in which some rules are applied to prevent confusion within clusters. A rule-based clustering can be applied to the given data. Then, next technique performs these rules with applying ontology-based semantics. And the last and basic technique changes presumed ontology and then rules will be applied on clusters. The results show that the derived clusters from the information provided within them were very similar and also very different from other clusters and, sequentially, a significant reduction in the k-distance of these clusters is also occurred and the correlation is increased.

After the talk, in a smaller group, we will continue the discussion, with a second and third part:

The second part will be dedicated to review of new research works and some various applications of such approaches. For example, some papers present the state of the art of this field such that different classes of approaches are covered such as linguistic, statistical, and machine learning including deep-learning-based approaches. Some relevant solutions are developed by reviewed papers which offer strategies and built-in methodologies for ontology learning. Ontologies place in the core of the semantic web and also very applicable resources for many artificial intelligence applications. Ontology learning, as a research area, proposes techniques to automate several tasks of the ontology construction process to simplify the boring work of manually building ontologies. Data analysis of biomedical approaches will be described as another application of such methods. Most researchers often tolerate severe logistical and technical challenges to discover, query, and merge heterogeneous data and knowledge from multiple sources. To cope with these challenges, the community has experimented with Semantic Web and linked data technologies to develop the Life Sciences Linked Open Data (LSLOD) cloud. In the reviewed paper, schemas from more than 80 biomedical linked open data sources are extracted into an LSLOD schema graph and conduct an empirical meta-analysis to appraise the extent of semantic heterogeneity across the LSLOD cloud. The other example is a paper which applied an extended version of the research paper “Ontology based crime investigation process” which deals with the working principle and the construction of an ontology based extensively on organized crime. Ontologies on various domains are made in which their assistance has been widely identified. They are being created for providing basis for allocation of knowledge in a domain and imparting reasonable information. In the reviewed paper the structure of the ontology is described and validated via an online ontology evaluating tool.

Finally, the third part will be my suggestions about the topic of my PhD thesis. Some related topics such as “Using ontology for geometry processing” or “Machine learning for chemical ontology classification” and etc. However, the exact choice for the topic will be dependent on the real requirements of the research team. I would prefer to work on topics and subjects that could help your team.