wchmm_add_word: CDSET phoneme exist in monophone?

[bayesrule]

Running own acoustic model + dictionary with given ENVR-v5.3 language models. Everything were fine before building HMM lexicon tree. Got cmd line error message: "wchmm_add_word: CDSET phoneme exist in monophone?"

Full output:

STAT: include config: xxx.jconf Stat: para: parsing HTK Config file: xxx_config Warning: para: "SOURCEFORMAT" ignored (not supported, or irrelevant) Warning: para: TARGETKIND skipped (will be determined by AM header) Stat: para: SOURCERATE=1250 Stat: para: TARGETRATE=200000.0 Warning: para: "SAVECOMPRESSED" ignored (not supported, or irrelevant) Warning: para: "SAVEWITHCRC" ignored (not supported, or irrelevant) Stat: para: WINDOWSIZE=250000.0 Stat: para: USEHAMMING=T Stat: para: PREEMCOEF=0.97 Stat: para: NUMCHANS=9 Stat: para: CEPLIFTER=22 Warning: para: NUMCEPS skipped (will be determined by AM header) STAT: jconf successfully finalized STAT: *** loading AM00 _default Stat: init_phmm: Reading in HMM definition Stat: rdhmmdef: ascii format HMM definition Stat: rdhmmdef: limit check passed Stat: check_hmm_restriction: an HMM with several arcs from initial state found: "sp" Stat: rdhmmdef: this HMM requires multipath handling at decoding Stat: rdhmmdef: no <SID> embedded Stat: rdhmmdef: assign SID by the order of appearance Stat: init_phmm: defined HMMs: 39425 Stat: init_phmm: loading ascii hmmlist Stat: init_phmm: logical names: 95243 in HMMList Stat: init_phmm: base phones: 47 used in logical Stat: init_phmm: finished reading HMM definitions STAT: m_fusion: force multipath HMM handling by user request STAT: making pseudo bi/mono-phone for IW-triphone Stat: hmm_lookup: 23 pseudo phones are added to logical HMM list STAT: *** AM00 _default loaded STAT: *** loading LM00 _default Stat: init_voca: read 76549 words Stat: init_ngram: reading in ARPA forward n-gram from ENVR-v5.3.bg Stat: ngram_read_arpa: this is 2-gram file Stat: ngram_read_arpa: reading 1-gram part... Stat: ngram_read_arpa: read 262145 1-gram entries Stat: ngram_read_arpa: reading 2-gram part... Stat: ngram_read_arpa: 2-gram read 0 (0%) Stat: ngram_read_arpa: 2-gram read 100000 (0%) Stat: ngram_read_arpa: 2-gram read 200000 (1%) Stat: ngram_read_arpa: 2-gram read 300000 (1%) Stat: ngram_read_arpa: 2-gram read 400000 (2%) Stat: ngram_read_arpa: 2-gram read 500000 (3%) Stat: ngram_read_arpa: 2-gram read 600000 (3%) ... ... Stat: ngram_read_arpa: 2-gram read 16200000 (98%) Stat: ngram_read_arpa: 2-gram read 16300000 (99%) Stat: ngram_read_arpa: 2-gram read 16380163 end Stat: init_ngram: found unknown word entry "" Stat: init_ngram: finished reading n-gram Stat: init_ngram: mapping dictonary words to n-gram entries Stat: init_ngram: finished word-to-ngram mapping STAT: *** LM00 _default loaded STAT: ------ STAT: All models are ready, go for final fusion STAT: [1] create MFCC extraction instance(s) STAT: *** create MFCC calculation modules from AM STAT: AM 0 _default: create a new module MFCC01 STAT: 1 MFCC modules created STAT: [2] create recognition processing instance(s) with AM and LM STAT: composing recognizer instance SR00 _default (AM00 _default, LM00 _default) WARNING: m_fusion: only -lmp or -lmp2 specified, LM weights may be unbalanced STAT: Building HMM lexicon tree

Any suggestions? Many thanks!

[Estalhun]

00readme-DNN.txt: "To prepare a model for DNN-HMM, note that the orders are important. The order of the output nodes in the DNN should be the order of HMM state definition id. If not, Julius won't work properly."

I think this is your problem.

[bayesrule]

@Estalhun Thank you so much! After hours of debugging, I have the same suspicion now. Although I'm not using DNN-HMM, but GMM instead, I think the cause is the same. The custom acoustic model that I'm using doesn't have <SID> included. Sad.