KenLM-docker

This repo provide the environment of execution KenLM. KenLM provide a n-gram language model (is not neural language model).

Requirements

• Docker

Usage

build docker image & train language model

$ make init-kenlm-docker  # wait a minute
$ make train-language-model N=3 \  # if N=3, training 3-gram language model
    INPUT_PATH=`a path of training target sentences` \
    OUTPUT_PATH=`a path where trained language model is output`

predict sentence probabilities with n-gram language model

• use kenlm python wrapper

$ make predict-with-language-model \
    OUTPUT_PATH=`a path of trained language model` \
    SENTENCE=`predicting target sentence (For example, "this is a pen .")`

References

• KenLM