SRILM Manual Pages

Programs

These are the top-level executables that are currently part of SRILM:

ngram-count

count N-grams and estimate language models

ngram-merge

merge N-gram counts

ngram

apply N-gram language models

ngram-class

induce word classes from N-gram statistics

disambig

disambiguate text tokens using an N-gram model

hidden-ngram

tag hidden events between words

nbest-lattice

rescore N-best lists and lattices

nbest-mix

interpolate N-best posterior probabilities

segment

segment text using N-gram language model

segment-nbest

rescore and segment N-best lists using N-gram language models

Utility Scripts

Additional tools implemented as scripts:

training-scripts

miscellaneous conveniences for language model training

lm-scripts

manipulate N-gram language models

ppl-scripts

manipulate perplexities

pfsg-scripts

create and manipulate finite-state networks

nbest-scripts

rescore and evaluate N-best lists

File Formats

Some of the data formats used by SRILM:

ngram-format

ARPA backoff N-gram models

classes-format

Word class definitions

pfsg-format

Decipher(TM) probabilistic finite-state grammars

nbest-format

N-best hypotheses lists

LM Library Classes

These are some of the basic classes of the SRILM library. Note that this list is woefully incomplete, as this part of the documentation is largely yet to be written.

LM

Generic language model

Vocab

Vocabulary indexing for SRILM

Prob

Probabilities for SRILM

File

Wrapper for stdio streams

Speech Group Home Page
SRI Home Page