SoFar

Recommender systems are well known means for finding the right product in a plethora of opportunities. One approach of is to guide the user with questions to the right product. Depending on product attributes and categorizations, multiple questions can be asked to narrow down the amount suitable products for the current user.

This kind of recommendation process needs well-structured semantic models as well as additional information about which features and attributes are relevant to customers. SoFar aims at providing an automated feedback cycle enhancing such kinds of question-driven recommenders.

The main contribution of the interactive systems group is to provide means for extracting sentiments and ratings about products of a given product ontology from unstructured and semi-structured resources such as blogs or customer reviews. A framework for crawling and analyzing unstructured resources is being developed, with focus on flexibility in terms of integration of new analyzation and crawling mechanisms. Intelligent rules are being integrated that map the analyzation results to entities within given product ontologies to finally provide an enriched ontology through standard compliant sparql interfaces.

Further information

Related research topics

Contact

Jürgen Ziegler

Full Professor

Further contributors

Tim Hussein

Former team member

Werner Gaulke

Former team member

Publication

Semi-automatic generation of recommendation processes and their GUIs