Meeting Recorder Naming Conventions

The goal of the naming conventions is to have consistent, short, easy to use names for people, microphones, channels, etc. We also wanted each field to be fixed length, so sorting and listing is easier.

Meeting IDs

An individual meeting (also called a session) is labeled with an alphanumeric tag. For example, "Bmr002". The tag consists of three fields. The first field must be alphabetic and all uppercase, and should be one letter. It represents the location of the recording. The locations assigned so far are:

BUC Berkeley/ICSI
WUniversity of Washington

The second field must be alphabetic and all lowercase, and should have two letters. It represents the meeting type. The following tags are currently assigned:

TagMeeting Type
dbDatabase issues meeting
edEven Deeper Understanding weekly meeting
ibIBM collaboration meeting
mrMeeting Recorder weekly meeting
nsNetwork Services and Applications group meeting
rgSpeech Reading group
roRobustness weekly meeting
skSmartKom dialogs
srSRI collaboration meeting
uwUW collaboration meeting

Note that each site may assign their own IDs for meeting types.

The final field must be numeric, consisting of three digits (e.g. "004").


A problem with this convention is the limited number of locations (26), meeting types (676), and meeting instances (1000). If we do end up with too many, we can either break the fixed field criterion, or relabel with more characters. Breaking the fixed field criterion perhaps isn't too bad, since the mixed-case and alphanumeric distinction still allows easy disambiguation.

Adding more characters counters our desire for short tags. We could also create a subdirectories for each location, but then you need two bits of information to label a meeting - the location and a tag, rather than just one short tag.

Opinions, anyone?

Speaker IDs

Speaker tags consist of "m", "f", or "u" for male, female, and unknown respectively, followed by "e" for native English speaker or "n" for non-native English speaker, followed by three numbers. The numbers should be unique across all speakers (e.g. there is only one person with a speaker ID ending in 003). For example, "fn002" and "me005" are both legal speaker IDs.

We have developed an XML database containing speaker information, including contact information, education level, language, age, etc. Take a look at the online speaker form to see the sort of information we collect.

For more information about the format of the speaker database, please contact

Mic IDs

A microphone ID represents the type of microphone. It is separate from the transmission ID, which represents the method by which the signal gets to the recording equipment. A microphone ID consists of a letter and a number. The microphone IDs used at ICSI are:

s1Sony headset mic ECM-310BMP
s2Sony handheld mic WRT-807A
c1Crown headset mic CM 311 A/E
l1Sony lapel (lavalier) ECM-77BMP
p1Plantronics monaural headset mic (part number unknown)
a1Andrea monaural headset mic NC-50
u1Unknown monaural headset mic
u2Unknown earplug mic
c2Crown PZM desktop microphone
u3Unknown microphone in mockup PDA

Transmission IDs

The transmission ID represents the method by which the microphone is connected to the recording equipment. It is a letter followed by a number. The channel IDs used at ICSI are:

j1Wired jack (via jimlet through jimbox)
w1Wireless Sony transmitter/receivers (Sony MB-806A modular base, WRU-806A/64 UHF synthesized tuner modules, WRT-805A bodypack transmitter)
w2Wireless Sony transmitter/receiver integrated into unit (e.g. Sony handset with integrated transmitter)
u1Unknown wired type (e.g. PZMs)

Digit Audio Filenames

As an example of the use of the conventions, we present here the naming convention of the digit audio files. These files contain the audio of a person reading a digit string. The files are named as follows:



For example, Bmr005-me001-l1-w1-b-1207-o5101.wav is from a Meeting Recorder meeting recorded at ICSI, a male English speaker, using the Sony lapel mic with the Sony transmitter/receiver, on channel "b", and the digits read were "oh five one zero one".

Questions or comments to: