Language-based multimedia information retrieval

This paper describes various methods and approaches for language-based multimedia information retrieval, which have been developed in the projects POP-EYE and OLIVE and which will be developed further in the MUMIS project. All of these project aim at supporting automated indexing of video material by use of human language technologies. Thus, in contrast to image or sound-based retrieval methods, where both the query language and the indexing methods build on non-linguistic data, these methods attempt to exploit advanced text retrieval technologies for the retrieval of non-textual material. While POP-EYE was building on subtitles or captions as the prime language key for disclosing video fragments, OLIVE is making use of speech recognition to automatically derive transcriptions of the sound tracks, generating time-coded linguistic elements which then serve as the basis for text-based retrieval functionality.