"Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?" T.-S. Eliot (1888-1965)

Knowledge Relationship Discovery aims at supporting search, visualization and analysis of complex knowledge spaces and thus providing knowledge in a format appropriate for human information processing. A crucial point in Knowledge Relationship Discovery is the identification of relationships between information, since the semantic dimensions of knowledge spaces do not only emerge from the data but to a great extend from the relations between data items. Our vision is the integration of content-based and semantic technologies to bridge the semantic gap between knowledge spaces.

Main Focus

Working on knowledge repositories such as document management systems, as well as more recent kinds of knowledge aggregations, such as Wikis, we use bottom-up approaches to semantically enrich data. We focus on textual data from dynamically changing data sources due to its semantic richness and its relevancy to praxis. However, media types like video, audio and images are provided by our scientific partners. Especially in times of the Prosumer webs, when content changes every second, strategies handling these dynamics have to be developed. Using methods like text and knowledge mining, ontology learning and ontology population we aim at improving the quality of information for post-processing steps like search and retrieval. For instance, information extraction strategies allow extraction and disambiguation of persons and integration of this information into semantic structures, like e.g. ontologies.

Based on the semantic enrichment of data sources we work on semantic harmonization of the enriched sources. We aim at harmonizing concepts of different data sources and comprising changes in the semantics of concepts. We apply ontology alignment and graph matching methods to convert semantic structures among each others. Additionally, on artifact level content based comparison methods and similarity analyses are applied. These technologies can be of important impact especially in the field of Web 2.0 technologies in enterprises for providing single-point-of-access mechanisms. For instance, Wikis more and more replace search and knowledge management systems as entry points to the central enterprise knowledge. Using the above mentioned technologies different contents can be dynamically linked to Wiki contents and therefore connect all information sources in the enterprise with the central entry point.

After semantic enrichment and harmonization the data and the extracted relations should be provided to the user. In this context novel visual and explorative methods for visualizing different semantic relationships are employed.
Using techniques from the fields of knowledge visualization and visual analytics we try to support the user in the deriving new knowledge. For example, these techniques are used to identify new topics in a patents data set over time. Additionally we apply semantic enrichment and harmonization to improve the quality of retrieval in heterogeneous data sources. An important issue in all these areas is the development of feedback methods to validate and improve methods and algorithms.

The developed KnowMiner framework serves as a basis for application-oriented projects and research activities. This framework provides a broad range of functionalities for rapid and efficient development and evaluation of new methods and technologies. Further it supplies numerous algorithms for the above mentioned areas. Current work on the KnowMiner framework includes service-oriented interfaces as well as the implementation and optimization of the framework to handle large-scale data sources.

Scientific fields of excellence

• Semantic enrichment of information sources using knowledge mining methods
• Semantic integration of heterogeneous information sources via ontologies and graph mining methods
• Retrieval in heterogeneous knowledge repositories and visualization of knowledge relationships

Competencies

• Knowledge visualization and human-computer-interfaces in the context of visualization and navigation of complex knowledge spaces
• User studies and technical evaluation based on statistical methods
• Service-oriented architectures and distributed, data-driven information processing