Information Service Engineering
In order to develop and provide new information services, the research department Information Service Engineering investigates models and methods for efficient semantic indexing, aggregation, linking, and retrieval of comprehensive heterogeneous and distributed data sources. To this end, both statistic and linguistic analysis methods (Natural Language Processing) as well as machine learning in combination with symbolic logic and interference mechanisms are applied.
The activities focus on:
- Automatic analyses (text and multi-media) with special emphasis on high efficiency and result quality. Besides fundamental new approaches, a combination of already existing methods (e.g., linguistic methods) are persued for text mining as well as for machine learning based classification of images and videos.
- Semantic analyses based on the results of raw data analysis in combination with existing metadata in order to sustainably index and automatically re-use document content, with a focus on accuracy and fault tolerance. This encompasses the following relevant sub-disciplines: Integration of heterogeneous meta data and content indexing via Named Entity Recognition, Named Entity Linking and Common Entity Linking
- Efficient semantic annotation of multi-media documents dependent on the document modality with automated user assistance (semi-automated annotation) as well as context-sensitive visualization and re-use of user generated annotations.
- Synchronization and integration of distributed heterogeneous knowledge bases with respect to provenance, reliability, diversity, multilingualism, and topicality (real-time).
- Semantic search, i.e., based on the results achieved via semantic analysis or after semantic annotation, the obtained semantic information is used to improve the retrieval process with respect to the completeness and accuracy of the search results.
- Explorative search and intelligent recommendation systems. There is a smooth transition from semantic search to explorative search and recommendation systems. While with semantic search results are generated based on semantic similarity, explorative search considers furtherrelationships among documents and their meaningful content elements.