Due to the rapidly increasing amounts of multimedia information
available on the web and in personal archives, good search algorithms
are becoming increasingly important. For a long time, there was a
tendency to treat different modalities, e.g. text, video and image
retrieval, separately. However, these modalities of information often
occur together, for example as tags associated with photographs and
videos, scripts associated with films, etc. This talk presents an
overview of recent techniques to search multiple modalities
simultaneusly, and to take advantage of the additional information
available from multiple modalities. When developing search algorithms,
it is important to be able to evaluate the quality of the returned
results. An overview of approaches to evaluating information retrieval,
some examples of evaluation campaigns, as well as shortcomings of
current approaches is presented.
|