Japanese Dictation Toolkit   - 1999 version -

Copyright (c)	1997-2000 Information-technology Promotion Agency (IPA), Japan
		http://winnie.kuis.kyoto-u.ac.jp/dictation/

				May 12, 2000


CONTENTS OF CD-ROM

doc/		Document
license		License Agreement
register	User Registration Form
julius/		Recognition Program (Julius)
phone_m/	Acoustic Model
lang_m/		Language Model
morph/		Morphological Analysis Tools
tools/		Tools
testrun/	Sample Program and Data


./doc:
guide.{pdf|ps}		General guide (Japanese)
guide_e.{pdf|ps}	General guide (English)
lvcsr.{pdf|ps}		Recognition program description (Japanese)
phone_m.{pdf|ps}	Acoustic model description (Japanese)
lang_m.{pdf|ps}		Language model description (Japanese)
morph.{pdf|ps}		Morphological analysis program description (Japanese)


./julius:
julius-3.1/		Recognition program Julius 3.1 source code
bin/			Recognition program Julius 3.1 binary code


./phone_m:
model/			HMM files
parms/			HMM parameters and phone inventory
tools/			Tools used for HMM training

./phone_m/model:
monof/			Monophone (CI) model
s1000/			Triphone model of 1000 states
s2000/			Triphone model of 2000 states
s3000/			Triphone model of 3000 states
PTM/			Phonetic Tied-Mixture (PTM) model

Each directory has a following structure
	mix4/			4-mixture model
	mix8/			8-mixture model
	mix16/			16-mixture model
		male/			male model
		female/			female model
		gid/			gender independent model

./phone_m/parms:
config.mfcc		acoustic analysis condition
config.train		HMM training condition
monophones		phone inventory
logicalTri.added	triphone inventory (logical -> physical mapping)
physicalTri		physical triphone inventory
tdc.hed			parameters for decision tree clustering
train,female.scp	training samples for female model
train,gid.scp		training samples for gender independent model
train,male.scp		training samples for male model


./lang_m:
5k/			5K-vocabulary model
20k/			20K-vocabulary model
60k/			60K-vocabulary model
LMcompress/		Language model compression program

./lang_m/5k:
5k.htkdic			5K lexicon with baseform
5k.vocab			5K lexicon without baseform
75.5k.0.wit.binlm.gz		forward bigram (CMU-BIN, 75-month, cutoff-0)
75.5k.1.wit.binlm.gz		forward bigram (CMU-BIN, 75-month, cutoff-1)
75.rev.5k.0-0.wit.binlm.gz	backward trigram (CMU-BIN, 75-month, cutoff-0-0)
75.rev.5k.1-1.10p.wit.arpa.gz	backward trigram (ASCII, 75-month, cutoff-1-1, compress-10%)
75.rev.5k.1-1.wit.binlm.gz	backward trigram (CMU-BIN, 75-month, cutoff-1-1)
bingram_for_julius/		LMs binary-formatted for Julius

./lang_m/20k:	
20k.htkdic			20K lexicon with baseform
20k.vocab			20K lexicon without baseform
45.rev.wit.1-1.10p.arpa.gz	backward trigram (ASCII, 45-month, cutoff-1-1, compress-10%)
45.rev.wit.1-1.binlm.gz		backward trigram (CMU-BIN, 45-month, cutoff-1-1)
45.rev.wit.4-4.binlm.gz		backward trigram (CMU-BIN, 45-month, cutoff-4-4)
45.wit.1.binlm.gz		forward bigram (CMU-BIN, 45-month, cutoff-1)
45.wit.4.binlm.gz		forward bigram (CMU-BIN, 45-month, cutoff-4)
75.rev.wit.1-1.10p.arpa.gz	backward trigram (ASCII, 75-month, cutoff-1-1, compress-10%)
75.rev.wit.1-1.binlm.gz		backward trigram (CMU-BIN, 75-month, cutoff-1-1)
75.rev.wit.4-4.binlm.gz		backward trigram (CMU-BIN, 75-month, cutoff-4-4)
75.wit.1.binlm.gz		forward bigram (CMU-BIN, 75-month, cutoff-1)
75.wit.4.binlm.gz		forward bigram (CMU-BIN, 75-month, cutoff-4)
bingram_for_julius/		LMs binary-formatted for Julius

./lang_m/60k:	
75.60k.htkdic			60K lexicon with baseform
75.60k.vocab			20K lexicon without baseform
75.60k.rev.wit.1-1.10p.arpa.gz	backward trigram (ASCII, 75-month, cutoff-1-1, compress-10%)
75.60k.rev.wit.1-1.binlm.gz	backward trigram (CMU-BIN, 75-month, cutoff-1-1)
75.60k.wit.1.binlm.gz		forward bigram (CMU-BIN, 75-month, cutoff-1)
75.60k.wit.1-1.10p.arpa.gz	forward trigram (ASCII, 75-month, cutoff-1-1, compress-10%)
75.60k.wit.1-1.binlm.gz		forward trigram (CMU-BIN, 75-month, cutoff-1-1)
bingram_for_julius/		LMs binary-formatted for Julius


./morph:
guideline		Guideline for pronunciation labeling
chasen-2.02/		Morphological analysis program ChaSen 2.0 source code
chawan-2.06/		Post-processor ChaWan 2.0 source code
postprocess-1.22/	Post-processor for exceptional expressions
suuzi_syori-1.0/	Post-processor for numerical expressions
jnas-ruby-patch/	Enhancement for JNAS corpus


./tools:
scoring/		Recognition result scoring tool
vocab2htkdic/		Lexicon tool (ChaSen + CMU-TK vocab. -> HTK lexicon)
wav2mfcc/		Acoustic analysis program (speech wave -> MFCC)
adapt-1.0/		Speaker adaptation program


./testrun:
go.sh			Script to test-run
jconf_best		Run-time configuration for Julius (accurate version)
jconf_fast		Run-time configuration for Julius (efficient version)
jconf_light		Run-time configuration for Julius (light version)
sample/			Sample speech file
sample_log/		Sample log file of Julius


AUTHORS

Kiyohiro Shikano	(Nara Inst. of Sci. & Tech.)
Tatsuya Kawahara	(Kyoto Univ.)
Kazuya Takeda		(Nagoya Univ.)
Katsunobu Itou		(ETL)
Akinori Ito		(Yamagata Univ.)
Takehito Utsuro		(Nara Inst. of Sci. & Tech.)
Tetsunori Kobayashi	(Waseda Univ.)
Shigeki Sagayama	(Japan Advanced Inst. of Sci. & Tech.)
Nobuaki Minematsu	(Toyohashi Univ. of Tech.)
Mikio Yamamoto		(Tsukuba Univ.)
Atsushi Yamada		(ASTEM)
Akinobu Lee		(Kyoto Univ.)