Concept Retrieval in PF/Tijah
Current multimedia search often falls back to retrieval in concept annotated collections. Specialized semantic concept detectors exploit low level features to annotate each resource of a collection with a score indicating whether a semantic concept is present in the resource.
They produce large amounts of textual meta-data that can be used later for retrieval. Searching in these meta-data collections differs from standard text retrieval, since detector scores should be taken into account.
Henning Rode shows in this talk how we address the work with concept collections in the framework of PF/Tijah: "I will finish the talk by explaining why the current 'more technical' changes might not be enough to fully solve the issue of concept search."
Location: CWI, Room C001
Presenter: Henning Rode
Please consult the TTT schedule for future talks at:
