MMIE training Duplicate base phone problem

[marekl]

Hi Nicolay,

I'm still unable to make MMIE working, after applying cepwin patch, I got
duplication error

INFO:   Initialization of the log add table
INFO:   Log-Add table size = 29356 x 2 >> 0
INFO:   
INFO: feat.c(855): Initializing feature stream to type: '1s_c', ceplen=13, CMN='current', VARNORM='no', AGC='none'
INFO: cmn.c(142): mean[0]= 12.00, mean[1..12]= 0.0
INFO:   Reading Feature Space Transform from: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/feature_transform
INFO:   Reading HMM in Sphinx 3 Model format
INFO:   Model Definition File: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/mdef
INFO:   Mean File: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/means
INFO:   Variance File: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/variances
INFO:   Mixture Weight File: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/mixture_weights
INFO:   Transition Matrices File: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/transition_matrices
INFO: mdef.c(683): Reading model definition: /home/monika/workspace/training/model_parameters/boolware.cd_cont_8000/mdef
FATAL_ERROR: "mdef.c", line 496: Duplicate base phone:    dz   -   - -    n/a    9     45     46     47     48     49 N

I have phonemes like dz and dZ in phone list which causes the following lines
in mdef

dZ   -   - -    n/a    8     40     41     42     43     44 N
dz   -   - -    n/a    9     45     46     47     48     49 N

It seems that Sphinx3 is case insensitive with regard to phonemes, am I right?
This is a problem for me as my phonemes are in consistence with SAMPA standard
and I would prefer not to change this. Is there any chance to work this
around?

[Nickolay V. Shmyrev]

Hello Marek

Change this line in sphinx3/src/libdecoder/libam/mdef.c

    m->ciphone_ht = hash_table_new(n_ci, 1);  /* With case-insensitive string names *//* freed in mdef_free */

to

    m->ciphone_ht = hash_table_new(n_ci, 0);  /* With case-sensitive string names *//* freed in mdef_free */

[marekl]

Thank you very much Nicolay, mdef has been loaded