Occasionally referred to as audio indexing, audio mining is a computerized task involving the processing of an audio file, extracting the dialog and creating a textual transcript, and searching the transcript for certain words or phrases. Considering the amount of audio content on the Internet and other sources, it is clear that audio mining is a growing technology.
To get an idea of what audio mining is and how it can be used, people can read this article from the Cutter Consortium (1
). It lists six broad areas that can benefit from using the technology and briefly discusses each one. A more detailed introduction is offered on the Leavitt Communications Web site (2
). This article delves into how audio mining works by giving a basic technical understanding of the process. A new method of searching an audio file, dubbed the "phonetic search engine," is compared to traditional methods in this white paper (3
). A publication from the Compaq Cambridge Research Laboratory (4
) discusses ways of collecting and analyzing information from an audio file. It also mentions SpeechBot, a Web-based tool for multimedia retrieval. Several papers can be downloaded from the home page of a research project studying the National Gallery of the Spoken Word (5
). The repository is comprised of massive historical audio content, and the team at the University of Colorado is investigating phrase recognition to index the data. Have you ever had a tune stuck in your head, but not known the name of the artist or song title? The Musical Audio-Mining project (6
) is working on ways to search for information about a song simply by humming part of it. Audio mining can also be used in the War on Terrorism, as is described in this article of Federal Computer Week (7
). Massive amounts of recorded phone conversations are intercepted by the government each day, and audio mining would be an efficient way to sort through irrelevant material and catch suspicious activity. The World Wide Web Consortium released this draft of the Voice Extensible Markup Language (8
), which could have applications for the audio mining community.
(no comments available yet for this resource)