Hi All,
I'd like to have your help to set up Simon using a French model.
I have been trying a lot of things for 1 week but can't make it work...

Here is the last procedure I did:
-> Configure Simon to use a Sphinx base model by specifiying 'mdef', 'means', ... files location, from here : http://sourceforge.net/projects/cmusphi ... c%20Model/
-> Import a Sphinx french dictionnary as Shadow dictionnary from here :
http://sourceforge.net/projects/cmusphi ... e%20Model/
-> Add 36 words in the active dictionnary and record my voice on them.

However when Simon compiles the model I get the following error:

Code: Select all

    /usr/bin/sphinxtrain -t default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7} setup
    Sphinxtrain path: /usr/lib/sphinxtrain
    Sphinxtrain binaries path: /usr/lib/sphinxtrain/sphinxtrain
    Setting up the database default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}

    /usr/bin/sphinxtrain run
    MODULE: 000 Computing feature from audio files
    Feature extraction is done
    MODULE: 00 verify training files
    Phase 1: Checking to see if the dict and filler dict agrees with the phonelist file.
    Found 39 words using 34 phones
    WARNING: This phone (au) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (bb) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (eu) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (in) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (ou) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (un) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (uu) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (uy) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    WARNING: This phone (ww) occurs in the dictionary (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.dic), but not in the phonelist (/tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/etc/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.phone)
    Phase 2: Checking to make sure there are not duplicate entries in the dictionary
    Phase 3: Check general format for the fileids file; utterance length (must be positive); files exist
    Phase 4: Checking number of lines in the transcript file should match lines in fileids file
    Phase 5: Determine amount of training data, see if n_tied_states seems reasonable.
    Estimated Total Hours Training: 0.0261055555555556
    This is a small amount of data, no comment at this time
    Phase 6: Checking that all the words in the transcript are in the dictionary
    Words in dictionary: 36
    Words in filler dictionary: 3
    Phase 7: Checking that all the phones in the transcript are in the phonelist, and all phones in the phonelist appear at least once
    MODULE: 0000 train grapheme-to-phoneme model
    Skipped (set $CFG_G2P_MODEL = 'yes' to enable)
    Feature type is s2_4x which is 4 streams
    LDA/MLLT only has sense for single stream features, for example 1s_c_d_dd
    Skipping LDA training
    Feature type is s2_4x which is 4 streams
    LDA/MLLT only has sense for single stream features, for example 1s_c_d_dd
    Skipping MLLT training
    MODULE: 05 Vector Quantization
    WARNING: This step had 0 ERROR messages and 1 WARNING messages. Please check the log file for details.
    MODULE: 10 Training Context Independent models for forced alignment and VTLN
    Skipped: $ST::CFG_FORCEDALIGN set to 'no' in sphinx_train.cfg
    Skipped: $ST::CFG_VTLN set to 'no' in sphinx_train.cfg
    MODULE: 11 Force-aligning transcripts
    Skipped: $ST::CFG_FORCEDALIGN set to 'no' in sphinx_train.cfg
    MODULE: 12 Force-aligning data for VTLN
    Skipped: $ST::CFG_VTLN set to 'no' in sphinx_train.cfg
    MODULE: 20 Training Context Independent models
    Phase 1: Cleaning up directories:
    accumulator...logs...qmanager...models...
    Phase 2: Flat initialize
    Phase 3: Forward-Backward
    Baum-Welch iteration 1 Average log-likelihood -0.552195149949778
    Baum-Welch iteration 3 Average log-likelihood 8.46721911321567
    Baum-Welch iteration 5 Average log-likelihood 10.2656134309083
    Baum-Welch iteration 8 Average log-likelihood 11.0200071746305
    Training completed after 10 iterations
    MODULE: 30 Training Context Dependent models
    Phase 1: Cleaning up directories:
    accumulator...logs...qmanager...
    Phase 2: Initialization
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    Phase 3: Forward-Backward
    Training failed in iteration 1
    MODULE: 40 Build Trees
    Phase 1: Cleaning up old log files...
    Phase 2: Make Questions
    Phase 3: Tree building
    Processing each phone with each state
    Skipping SIL
    MODULE: 45 Prune Trees
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    Phase 1: Tree Pruning
    This step had 2 ERROR messages and 0 WARNING messages. Please check the log file for details.
    Phase 2: State Tying
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    MODULE: 50 Training Context dependent models
    Phase 1: Cleaning up directories:
    accumulator...logs...qmanager...
    Phase 2: Copy CI to CD initialize
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    Phase 3: Forward-Backward
    Training failed in iteration 1
    MODULE: 60 Lattice Generation
    Skipped: $ST::CFG_MMIE set to 'no' in sphinx_train.cfg
    MODULE: 61 Lattice Pruning
    Skipped: $ST::CFG_MMIE set to 'no' in sphinx_train.cfg
    MODULE: 62 Lattice Format Conversion
    Skipped: $ST::CFG_MMIE set to 'no' in sphinx_train.cfg
    MODULE: 65 MMIE Training
    Skipped: $ST::CFG_MMIE set to 'no' in sphinx_train.cfg
    MODULE: 90 deleted interpolation
    Phase 1: Cleaning up directories: logs...
    Phase 2: Doing interpolation...
    This step had 1 ERROR messages and 0 WARNING messages. Please check the log file for details.
    Phase 3: Dumping senones for PocketSphinx...
    This step had 2 ERROR messages and 0 WARNING messages. Please check the log file for details.
    MODULE: DECODE Decoding using models previously trained
    Aligning results to find error rate
    Sphinxtrain path: /usr/lib/sphinxtrain
    Sphinxtrain binaries path: /usr/lib/sphinxtrain/sphinxtrain
    Running the training

    Failed to copy /tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/model_architecture/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.200.mdef to /tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/model_parameters/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}.cd_semi_200_delinterp/mdef: Aucun fichier ou dossier de ce type at /usr/lib/sphinxtrain/scripts/90.deleted_interpolation/deleted_interpolation.pl line 110.
    Can't open /tmp/kde-user/simond/default/compile/sphinx/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}/result/default{22d71ca0-d638-4e88-814f-a4a5ecbc75a7}-1-1.match
    word_align.pl failed with error code 65280 at /usr/lib/sphinxtrain/scripts/decode/slave.pl line 173.


Can somebody help me with this problem ?
Thanks !

Hello,

does it work when you set Simon to use a "static base model" in the base model configuration?

Best regards,
Peter