Reachability Analysis of Graph modelled Collections

Nowadays, we have a large number of information sources from different modalities such as text,  images,  audio and  video. Information objects are not isolated anymore. They are frequently connected via metadata, semantic links, etc. This leads to various challenges in graph-based Information Retrieval. This talk is concerned with how much graph models help reaching relevant information objects, especially from different modalities. A framework is presented in which different features and relations between information objects are exploited to improve recall. The experiments are performed on the Image CLEF 2011 Wikipedia collection.  The results show that the combination of different facets is conductive to obtain higher reachability.