Fast browsing of found audio data

Abstract:

In this yearly seminar I will present my progress so far in my PhD
studies in the project TillTal. The aim is to make it possible to
utilise found data - data that was not recorded with the specific
purpose of being used in research. Typical examples of found resources
are political speeches, radio segments, interviews, audio-visual data
such as television and movies, music, recordings and transcriptions of
the NASA’s Apollo missions, and archived material in general. Adopting a
concept that we have come to call temporally disassembled audio we have
managed to produce promising results for browsing, and to some extent
annotating, large quantities of audio. I will present this work, demo a
tool that has been built based on these ideas, and briefly mention our
plans for future work. Welcome!