Prepare files for ctc_decoder Data Text to generate the LM and lexicon should be in data. Usage cd scripts bash decoder_files.sh Different n-gram models can be created by changing the DATA_PT and N_GRAM parameters in this script.