Community Research and Development Information Service - CORDIS

Speech recognition tool for broadcast material retrieval

The Media industry makes a substantial investment in archiving multimedia material that apart from its commercial value, by now has become a record of our cultural evolvement. Access to this archived material is now further facilitated with a multilingual indexing tool based on speech recognition.
Speech recognition tool for broadcast material retrieval
The high content value of multimedia archived material was, until recently accessed through processing of subtitles. Methods were developed for linking indexes to sections of video. The current project, termed OLIVE, will extend the current methods and introduce new ones for archiving retrieval. Using speech recognition a transcript will be produced from a broadcast soundtrack that will be time linked to the original recording. Generation of a concept index from this transcript will be used for retrieval. In addition, the availability of fuzzy matching will further enhance the retrieval capabilities by connecting phrases with close meaning. Queries in four different languages, German, French, English and Dutch, are supported.

Phrase indexing, Query translation and Speech recognition technologies allow the automatic production of indexes from the audio track of a radio or television programme to be used for retrieval. However, prior to this, the provision of bibliographic material, transcripts and video stills, allows the user to preview the material and decide on its usefulness, or relevance to the users.

A great variety of end users will benefit from the novelty of this multimedia application. The owners of broadcast material will be provided with new paths that will lead to greatly enhanced exploitation. Broadcasting companies, especially for the production of their documentary, scientific or educational programmes, will now be able to use previously existing, archived radio and video material.

Service providers such as, television production companies, or advertising companies will be able to search databases of video for sequences that could be used for new productions, or included in their existing productions. Researchers, scientists and journalists, to name a few, will gain access to this wealth of information that more dramatically than any other means, testifies to human endeavour.

Project URL: http://twentyone.tpd.tno.nl/olive/

Related information

Record Number: 80228 / Last updated on: 2005-09-18
Domain: IT, Telecommunications