Information Retrieval in the Audio Domain

摘要: Multimedia content in the form of non-textual data is becoming more and abundant. This chapter addresses problem adequate natural efficient access to this increasing amount available content, sometimes referred as bottleneck. We focus on special case spoken audio demonstrate how partly erroneous document information (e.g. from automatic speech recognizer transcripts) may be consistently incorporated into existing retrieval infrastructure. Experiments results a standard task are presented.