.TH PHONETISAURUS "1" "February 2013" "phonetisaurus 0.7.8" "User Commands"

.SH NAME

phonetisaurus\-align \- Dictionary aligner

.SH SYNOPSIS

\fBphonetisaurus\-align\fR \-\-input=\fIdictionary.bsf\fR \-\-ofile=\fItraining.corpus\fR [\fIOPTIONS\fR]

.SH DESCRIPTION

\fBphonetisaurus\-align\fR

This tool read an input dictionary and produce an aligned corpus that
can be used to train a model for Grapheme\-to\-Phoneme conversion.

.SH INPUT FORMAT

The input format is a two columns plain-text file. The first column is
supposed to contain a graphemes sequence (e.g., the orthographic form
of a word). The second column is supposed to contain the corresponding
phonemes sequence.

By default the two columns are separated by a TAB character (it is
possible to change the separator using the \fB\-\-delim\fR option),
each character of the first column is supposed to be a grapheme (it is
possible to specify a grapheme separator using \fB\-\-seq1_sep\fR),
phonemes in the second column are separated by spaces (it is possible
to change the phoneme separator using \fB\-\-seq2_sep\fR).

Input example:

ABBREVIATE      AH B R IY V IY EY T

.SH OPTIONS

.HP
\fB\-\-help\fR=<\fIbool\fR> (default: false)
.IP
show usage information
.HP
\fB\-\-helpshort\fR=<\fIbool\fR> (default: false)
.IP
show brief usage information
.HP
\fB\-\-tmpdir\fR=<\fIstring\fR> (default: "/tmp/")
.IP
temporary directory
.HP
\fB\-\-v\fR=<\fIint32\fR> (default: 0)
.IP
verbose level
.HP
\fB\-\-fst_align\fR=<\fIbool\fR> (default: false)
.IP
Write FST data aligned where appropriate
.HP
\fB\-\-fst_default_cache_gc\fR=<\fIbool\fR> (default: true)
.IP
Enable garbage collection of cache
.HP
\fB\-\-fst_default_cache_gc_limit\fR=<\fIint64\fR> (default: 1048576)
.IP
Cache byte size that triggers garbage collection
.HP
\fB\-\-fst_verify_properties\fR=<\fIbool\fR> (default: false)
.IP
Verify fst properties queried by TestProperties
.HP
\fB\-\-fst_weight_parentheses\fR=<\fIstring\fR> (default: "")
.IP
Characters enclosing the first weight of a printed composite weight (e.g. pair weight, tuple weight and derived classes) to ensure proper I/O of nested composite weights; must have size 0 (none) or 2 (open and close parenthesis)
.HP
\fB\-\-fst_weight_separator\fR=<\fIstring\fR> (default: "")
.IP
Character separator between printed composite weights; must be a single character
.HP
\fB\-\-save_relabel_ipairs\fR=<\fIstring\fR> (default: "")
.IP
Save input relabel pairs to file
.HP
\fB\-\-save_relabel_opairs\fR=<\fIstring\fR> (default: "")
.IP
Save output relabel pairs to file
\fB\-\-delim\fR=<\fIstring\fR> (default: "	")
.IP
Delimiter used to separate input and output tokens.
.HP
\fB\-\-eps\fR=<\fIstring\fR> (default: "<eps>")
.IP
Epsilon symbol.
.HP
\fB\-\-fb\fR=<\fIbool\fR> (default: false)
.IP
Use forward\-backward pruning for the alignment lattices.
.HP
\fB\-\-input\fR=<\fIstring\fR> (default: "")
.IP
Two\-column input file to align.
.HP
\fB\-\-iter\fR=<\fIint32\fR> (default: 11)
.IP
Maximum number of EM iterations to perform.
.HP
\fB\-\-lattice\fR=<\fIbool\fR> (default: false)
.IP
Write out the alignment lattices as an fst archive (.far).
.HP
\fB\-\-model\fR=<\fIbool\fR> (default: true)
.IP
Load a pre\-trained model for use.
.HP
\fB\-\-mbr\fR=<\fIbool\fR> (default: false)
.IP
Use the LMBR decoder (not yet implemented).
.HP
\fB\-\-model_file\fR=<\fIstring\fR> (default: "")
.IP
FST\-format alignment model to load.
.HP
\fB\-\-nbest\fR=<\fIint32\fR> (default: 1)
.IP
Output the N\-best alignments given the model.
.HP
\fB\-\-ofile\fR=<\fIstring\fR> (default: "")
.IP
Output file to write the aligned dictionary to.
.HP
\fB\-\-penalize\fR=<\fIbool\fR> (default: true)
.IP
Penalize scores.
.HP
\fB\-\-penalize_em\fR=<\fIbool\fR> (default: false)
.IP
Penalize links during EM training.
.HP
\fB\-\-pthresh\fR=<\fIdouble\fR> (default: \-99)
.HP
Pruning threshold.  Use to prune unlikely N\-best candidates when using multiple alignments.
.HP
\fB\-\-restrict\fR=<\fIbool\fR> (default: true)
.IP
Restrict links to M\-1, 1\-N during initialization.
.HP
\fB\-\-s1_char_delim\fR=<\fIstring\fR> (default: "")
.IP
Sequence one input delimiter.
.HP
\fB\-\-s1s2_sep\fR=<\fIstring\fR> (default: "}")
.IP
Token used to separate input\-output subsequences in the g2p model.
.HP
\fB\-\-s2_char_delim\fR=<\fIstring\fR> (default: " ")
.IP
Sequence two input delimiter.
.HP
\fB\-\-seq1_del\fR=<\fIbool\fR> (default: true)
.IP
Allow deletions in sequence one.
.HP
\fB\-\-seq1_max\fR=<\fIint32\fR> (default: 2)
.IP
Maximum subsequence length for sequence one.
.HP
\fB\-\-seq1_sep\fR=<\fIstring\fR> (default: "|")
.IP
Multi\-token separator for input tokens.
.HP
\fB\-\-seq2_del\fR=<\fIbool\fR> (default: true)
.IP
Allow deletions in sequence two.
.HP
\fB\-\-seq2_max\fR=<\fIint32\fR> (default: 2)
.IP
Maximum subsequence length for sequence two.
.HP
\fB\-\-seq2_sep\fR=<\fIstring\fR> (default: "|")
.IP
Multi\-token separator for output tokens.
.HP
\fB\-\-skip\fR=<\fIstring\fR> (default: "_")
.IP
Skip token used to represent null transitions.  Distinct from epsilon.
.HP
\fB\-\-thresh\fR=<\fIdouble\fR> (default: 1e\-10)
.IP
Delta threshold for EM training termination.
.HP
\fB\-\-write_model\fR=<\fIstring\fR> (default: "")
.IP
Write out the alignment model in OpenFst format to filename.
.HP
\fB\-\-fst_compat_symbols\fR=<\fIbool\fR> (default: true)
.IP
Require symbol tables to match when appropriate
.HP
\fB\-\-fst_field_separator\fR=<\fIstring\fR> (default: "      ")
.IP
Set of characters used as a separator between printed fields
.HP
\fB\-\-fst_error_fatal\fR=<\fIbool\fR> (default: true)
.IP
FST errors are fatal; o.w. return objects flagged as bad:
e.g., FSTs \- kError prop. true, FST weights \- not  a Member()