Documentation for the 1998 Speaker Recognition Evaluation Test-Set ------------------------------------------------------------------ The 1998 Speaker Recognition Evaluation Test-Set is contained in a 6-CDrom set. -------------------------------------------------- NIST SPEECH DISC NUMBER || Top Level Directory -------------------------------------------------- r46_1_1 || sid98e1f r46_2_1 || sid98e2f r46_3_1 || sid98e3f r46_4_1 || sid98e1m r46_5_1 || sid98e2m r46_6_1 || sid98e3m -------------------------------------------------- The Evaluation Plan (which provides detailed documentation used in running the evaluation) may be found on each disc in the documentation directory (i.e sid98e1f/doc) It exists in two forms, Postscript "spk98v13.ps" and HTML "spk98v13.htm". This README file is intended to provide specific information about the 6-disc CDrom test-set. ------------- Training Data ------------- Female: The female training data is located on two discs. The first disc (r46_1_1) ------------------------ One-Session training: sid98e1f/train/s1a/ sid98e1f/train/s1b/ Two-Session training: sid98e1f/train/s1a/ sid98e1f/train/s2a/ The second disc (r46_2_1) ------------------------ Two-Session-full training: sid98e2f/train/s1r/ sid98e2f/train/s2r/ Note, this disc contains the "additional" data to be used for the two-session-full test. It is to be used in conjunction with the training data on the first disc. ------------------------------------------------------------------------- WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING There are some empty sphere files for 16 female speakers in the s1r train directory. The following is a list of these files. (They actually contain one byte of data, in order for SPHERE functions to work properly.) These files are located at: ./sid98e2f/train/s1r/ 3013.sph 3105.sph 3187.sph 3212.sph 3290.sph 3356.sph 3387.sph 3439.sph 3493.sph 3526.sph 3571.sph 3576.sph 3637.sph 3641.sph 3707.sph 3735.sph ------------------------------------------------------------------------- Male: The male training data is located on two discs. The first disc (r46_4_1) ------------------------ One-Session training: sid98e1m/train/s1a/ sid98e1m/train/s1b/ Two-Session training: sid98e1m/train/s1a/ sid98e1m/train/s2a/ The second disc (r46_5_1) ------------------------ Two-Session-full training: sid98e2m/train/s1r/ sid98e2m/train/s2r/ Note, this disc contains the "additional" data to be used for the two-session-full test. It is to be used in conjunction with the training data on the first disc. ------------------------------------------------------------------------- WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING WARNING There are some empty sphere files for 20 male speakers in the s1r train directory. The following is a list of these files. (They actually contain one byte of data, in order for SPHERE functions to work properly.) These files are located at: ./sid98e2m/train/s1r/ 3038.sph 3068.sph 3073.sph 3081.sph 3137.sph 3156.sph 3243.sph 3293.sph 3312.sph 3419.sph 3429.sph 3435.sph 3440.sph 3575.sph 3718.sph 3746.sph 3821.sph 3882.sph 3890.sph 3894.sph ------------------------------------------------------------------------- ------------------------------------------------------------------------- --------- Test Data --------- Female: The female test data is located on two discs. The first disc (r46_2_1) ------------------------ 3-Second Test Segments: sid98e2f/test/3/ 10-Second Test Segments: sid98e2f/test/10/ The second disc (r46_3_1) ------------------------ 30-Second Test Segments: sid98e3f/test/30/ There are three index files in each test duration directory, which are to be used for the tests. (See the Evaluation Plan) Male: The male test data is located on two discs. The first disc (r46_5_1) ------------------------ 3-Second Test Segments: sid98e2m/test/3/ 10-Second Test Segments: sid98e2m/test/10/ The second disc (r46_6_1) ------------------------ 30-Second Test Segments: sid98e3m/test/30/ There are three index files in each test duration directory, which are to be used for the tests. (See the Evaluation Plan) ------------------- Extra Test Segments ------------------- Female: The female test data that has been created but is not being used for this evaluation, is located on one disc. There is an index file "unused.ndx", located in each test duration directory, which simply lists these files. Disc (r46_1_1) -------------- 3-Second Test Segments: sid98e1f/test/3/ 10-Second Test Segments: sid98e1f/test/10/ 30-Second Test Segments: sid98e1f/test/30/ Male: The male test data that has been created but is not being used for this evaluation, is located on one disc. There is an index file "unused.ndx", located in each test duration directory, which simply lists these files. Disc (r46_4_1) -------------- 3-Second Test Segments: sid98e1m/test/3/ 10-Second Test Segments: sid98e1m/test/10/ 30-Second Test Segments: sid98e1m/test/30/