MOTOROLA

The main object of the project is to contribute to Motorola’s Recommender System Development, for use in a number of areas, such as content recommendation and personalized advertising. Most existing techniques for recommendation and content-based advertising rely on the exploitation of lexical relationships between the content a user has consumed and the content to be recommended. More advanced techniques often use semantic representations, such as domain-specific ontologies, which extend the power of recommender systems, but such systems often have two deficiencies, they can be large and inefficient, and the ontologies themselves are typically hand-crafted, which greatly limits their utility. This project aims to solve both of these shortcomings by building upon previous CERTH/ITI and Motorola research to research and develop Recommender Systems which derive from the automatic induction of ontologies from data, and to further research how such automatically generated semantic structures can be stored and processed efficiently, and merged and amended with human-crafted semantic structures when available.