Scroll to navigation

INTERPOLATE-NGRAM(1) User Commands INTERPOLATE-NGRAM(1)

NAME

interpolate-ngram - interpolates n-gram language models

SYNOPSIS

interpolate-ngram [Options]

DESCRIPTION

Interpolates multiple n-gram models by computing appropriate interpolation weights from optional features and constructing a statically interpolated n-gram model. Parameters can be optionally tuned to optimize development set performance.

Filename argument can be an ASCII file, a compressed file (ending in .Z or .gz), or '-' to indicate stdin/stdout.

OPTIONS

Print this message.
Set verbosity level.
Default: 1
Set the n-gram order of the estimated LM.
Default: 3
Fix the vocab to only words from the specified file.
Replace all out of vocab words with <unk>.
Default: false
Interpolate specified LM files.
Interpolate models trained from text files.
Interpolate models trained from counts files.
Specify smoothing algorithms.
Default: ModKN
Specify n-gram weighting features.
Specify interpolation mode.
Default: LI
Specify interpolation features.
Tie parameters across n-gram order.
Default: true
Tie parameters across LM components.
Default: false
Set initial model params.
Specify optimization algorithm.
Default: LBFGS
Tune params to minimize dev set perplexity.
Tune params to minimize lattice word error rate.
Tune params to minimize lattice margin.
Write LM/counts files in binary format.
Default: false
Write tuned model params to file.
Write LM vocab to file.
Write ARPA backoff LM to file.
Compute test set perplexity.
Compute test set lattice word error rate.
Compute test set lattice margin.

SEE ALSO

estimate-ngram(1), evaluate-ngram(1)

January 2013 MITLM