Data Organization and Naming Schema

The sound files contained in the corpus are .WAV files sampled at 44100 Hz with a resolution of 16 bits. The content is organized according to the following hierarchy:

/data/[SpeakingStyle]/[SpeakerID]/[filename].wav

where:

[SpeakingStyle]	:= fast|retell|rsi|solo|sync|whsp
[SpeakerID] 	:= the string identifying the speaker

the label sync indicates SYNCHRONOUS recordings, while whsp indicates WHISPERed recordings. In order to maximise the readability of the [filename], it is constructed as follows:

[filename] := [SpeakerID]_[TextID]_[SpeakingStyle]_[TargetID]

where:

[SpeakerID] := the string identifying the speaker
[TextID] := the code uniquely identifying the text
[SpeakingStyle]	:= fast|retell|rsi|solo|sync|whsp
[TargetID] := id of the target/co-speaker

The tag TargetID is used only in the relevant speaking styles: sync and rsi. As conditions SOLO, SYNCHRONOUS and RETELL were recorded in a studio, additional tracks recorded on different microphones are available if needed. These additional tracks are not being released in this initial version of the corpus, but may be obtained by applying to the authors.