CMU Dictionary Adaptation to Bahasa Indonesia Lexicon Building

While creating lexicon in voice recognition in Bahasa Indonesia, we need to define the phoneme set by ourselves since there are not such a widely used standard in Bahasa Indonesia. Instead of creating new definition, there is an idea to adapt from existing phoneme set. Obtained from Kaldi resources, we can adapt the phoneme set from English issued by Carnegie Mellon University (CMU Dictionary) which contains 134,000 words.

Bahasa Indonesia is quite simple^{look here also} as in major case the pronunciation and written letter are the same compared to English. Thus, it is not a tedious work to start building lexicon based on CMU Dictionary although we need to add new phonemes and leave some phonemes.

Table 1 CMU Dictionary Phoneme Set

Phoneme	IPA Symbol	Indonesian Words Example	English Words Example
AA	ɑ	ternak, gembala	odd, balm
AE	æ	-	at, bat
AH	ʌ	ambil	hut, butt
AO	ɔ	bakpao	cow, story
AW	aʊ	saung	ought, bout
AY	aɪ	kait, senarai	hide, bite
B	b	lebah, beli	bee, buy
CH	tʃ	cuka, ceri	cheese, china
D	d	diam, duduk	dump, did
DH	ð	ridho	the, thy
EH	ɛ	enak, sepak	education, bet
ER	ɝ	bageur*, reueus*	hurt
EY	eɪ	-	ate, bait
F	ɾ	faedah, fana	fee, forest
G	g	gerbang, guna	green, gate
HH	h	hampar, unggah	he, hair
IH	ɪ	singgah, ikatan	it, implication
IY	i	-	eat, sheep
JH	dʒ	adiraja, keganjilan	genuine, jimmy
K	k	kenangan, batuk	key, camp
L	l	lingkaran, betul	luck, love
M	m	minum, temaram	mama, mine
N	n	naik, menikah	knee, nice
NG	ŋ	yang, ngengat	bank, sink
OW	oʊ	bongkar, bogor	oat, boat
OY	ɔɪ	-	toy, boy
P	p	pulsa, peluh	pulp, pen
R	ɹ	ranjau, rintangan	right, row
S	s	sakit, sayang	sea, sun
SH	ʃ	masyarakat, syaikh*	shine, she
T	t	tikung, timpa	tea, tone
TH	θ	rabiul tsani*	thug, theta
UH	ʊ	-	hood, book
UW	u	kuku, suku	two, coup
V	v	viral, vas	vee, vocal
W	w	wejangan, wayang	we, wide
Y	j	yakin, yoga	yam, yield
Z	z	zaman, zamrud	zoo, zee
ZH	ʒ	jangkrik, jerapah	seizure, pleasure

From table above, it’s clear that there are phonemes that are not (commonly) used in Bahasa Indonesia. Yet these phoneme set does not cover all Bahasa Indonesia lemma regarding to the root of Bahasa Indonesia which come from majorly Malay, Dutch, Arabic, Chinese, Javanese, and Sundanese. To create the things short, here is list of phonemes that needs to be added to CMU Dictionary for Bahasa Indonesia

Table 2 Addition Phoneme Set

Phoneme	IPA Symbol	Indonesian Words Example	Notes
NY	ɲ	kenyang, nyamuk	alveolo palatal
KH	x	kholifah	voiceless velar fricative
Q	ʔ	qurban
KX	ʕ	sa’at	voiced pharyngeal fricative
DL	zˁ	dhuhur	voiced alveolar sibilant with pharyngealization
GH	ɣ	ghaib	voiced velar fricative

sources :
[1] http://kaldi-asr.org/doc/examples.html
[2] http://www.speech.cs.cmu.edu/cgi-bin/cmudict
[3] https://en.wikipedia.org/wiki/ARPABET
[4] https://open-dict-data.github.io/ipa-lookup/ma/