The Voicemail Corpus - Part I transcriptions include the following
information in the specified format:


coughing, laughter, breath noise, inhaling, marked as <cough>,
<laughter>, <breath>, <inhale>.

clicks and beeps in the message marked as <click>, <beep>.

other disfluencies marked as one of the following categories
<AH>, <AHA>, <ERR>, <HMM>, <HUM>, <HA>, <HO>, <HUH>, <MMM>, <OH>,
<UH>, <UM>

if someone stammers and says 'thir-thirty', the corresponding
transcription would be 'thir- thirty'

no punctuation (for instance ; : or ,) used in the transcription

all scripts are upper-case

when transcribing times, the convention is A.M. and P.M.
instead of A M and P M (this would be the only case where a punctuation mark
is used)

hyphen used when transcribing words like E-mail and other compound
words such as give-me-a-call, thank-you, etc.

names in the message preceded with a ! sign.
This includes proper names, names of companies, days of week, and months. For
instance Apple Computer would be transcribed as !Apple !Computer.

numbers spelled out instead of using numbers, for instance 1997 would
be transcribed as nineteen ninety-seven.

when transcribing spelled out letters, for example, if someone spells out
the letters in a word, for example I B M, transcribed as !I.B.M.,
i.e. "." after every letter that is spelt out. 

if a segment of acoustic is incomprehensible, we made up a spelling that
sounded like the acoustic, rather than mark them as mumble words. these
non-words are marked with a ? or @ sign at the start of the spelling'