NAME

hmmscribe.tcl - Select pronunciation variants using Viterbi forced alignment.

AVAILABILITY

script/hmm_1.0

SYNOPSIS

 hmmscribe.tcl modelname [options]

PARAMETERS

modelname

HMM model used to generate forced aligned transcriptions (e.g. digit.1)

OPTIONS

-basename string

Alternative base model name (used to load male/female models which share the same HMM architecture and therefore the same model description files .rr .list)

-dictionary string [Default = dict.db]

Word pronunciation dictionary database. (see worddb.tcl)

-input string [Default = transcript]

Input word level transcriptions (Master Label File).

-output string [Default = alignment]

Output forced-aligned transcriptions (Master Label File).

-word boolean [Default = 1]

Return the word level alignment as a MLF .wrd file.

-phone boolean [Default = 0]

Return the phone level alignment as a MLF .phn file.

-state boolean [Default = 0]

Return the state level alignment as a MLF .ste file.

-config string

Read command line options (configuration info) from this file. The configuration file is in essence a Tcl script which sets the required internal variables.

 set config(scribe,dictionary) dict.db
 set config(scribe,input) transcript
 set config(scribe,output) alignment

These values will override the preset default values. All subsequent command line options will override the values specified by the configuration script. Command line parameters are specified using the param variable.

 set param(scribe,modelname) foo

Since command line parameters are typically not optional the user needs to specify the command line parameters as a single "-" character for the settings defined in the configuration script to take effect.

The configuration file is also used to specify the feature post-processing script.

 set config(feature,script) user.tcl

If this variable is not defined in the configuration file, then only the base features as saved in the feature cache are used during training.

-help

Provides a short description of the command line options.

DESCRIPTION

hmmscribe.tcl selects the word pronunciation variants as they appear in the word pronunciation dictionary.

The input word transcriptions are used to create a finite state grammar where each node or state in the grammar contains a word and its pronunciation variants. The standard Viterbi algorithm is then used to find the best possible path through the grammar, resulting in the selection of the pronunciation variants to fit the sentence.

The resulting output word transcriptions may then be used to generate the associated model transcriptions for HMM embedded training (see genmodel.tcl, hmmembed.tcl).

SEE ALSO

gendict.tcl, worddb.tcl, Master Label File

AUTHOR

Johan Schalkwyk
Center for Spoken Language Understanding
Oregon Graduate Institute of Science & Technology