The format of the key fields is as follows:

Field 1:  Segment Name

Field 2:  Nominal Test Duration

Field 3:  Actual Test Duration

Field 4:  Switchboard conversation from which the segment was
          extracted.

Field 5:  Speaker Identification Number

Field 6:  [. unused .]

Field 7:  Channel of Switchboard conversation from which the segment
          was extracted.

Field 8:  Gender,  "m" -> male   "f" -> female

Field 9:  target or non-target   "tar" -> targets
          Note:

               For this evaluation they are all targets.  The only
               reason this field would be marked differently, is if
               in the future we find some incorrectly labeled
               speakers.

Field 10: Handset relation to training
          SS --> Same handset as was used in training  (same_num)
          SD --> Different handset (phone number) same handset type
                 (electret or carbon) as was used in training.
          DD --> Different handset (phone number) different handset
                 type (electret or carbon) as was used in training.

Field 11: Test segment quality
          g --> good
          b --> bad
          r --> really bad

          Each "b" and "r" has a number attached.  These numbers are
          catagories of why the segment was marked as bad or really
          bad.

          Summary:      1:  Line Noise
                        2:  Cross talk
                        3:  Background music/vocal/animal sounds
                        4:  Background environmental kitchen/office
                            sounds
                        5:  [. unused .]
                        6:  Intermittent static
                        7:  Intermittent handset-handling noise
                        8:  continual Mic or Channel Distortion
                        9:  Continual breathing into the mic
                        10: Excessive amounts of silence/laughing
                            or any other non-speech noise

Field 12: Elapsed time of the conversation it took to get the speech
          segment.  From the beginning of the segment to the end,
          including silences.

Field 13: [. unused .]

Field 14: Age of target speaker

Field 15: State in which the target speaker was raised

Field 16: Handset type of test segment
          "elec" -> electret    "carb" -> carbon_button

Field 17: Average pitch of test segment.

Field 18: "s1a" training segment quality
          g --> good
          b --> bad
          r --> really bad

          same rules apply for the number as above.

Field 19: "s2a" training segment quality
          g --> good
          b --> bad
          r --> really bad

          same rules apply for the number as above.

Field 20: Handset type of training segments

Field 21: Average pitch of training segment "s1a"

Field 22: Average pitch of training segment "s1b"

Field 23: Average pitch of training segment "s1f"

Field 24: Average pitch of training segment "s2a"

Field 25: Average pitch of training segment "s2f"