Phonetic features of consonants and vowels — data_features

These are two dataframes that contain conventional phonetic features of the consonants and vowels used by CMU phonetic alphabet.

Usage

data_features_consonants

data_features_vowels

Format

An object of class tbl_df (inherits from tbl, data.frame) with 24 rows and 10 columns.

An object of class tbl_df (inherits from tbl, data.frame) with 17 rows and 12 columns.

Details

Most of the features are self-evident and definitional. For example, /p/ is the bilabial voiceless stop. For fuzzier features, I consulted the general IPA chart and the Wikipedia page on English phonology. These issues included things like: what are the lax vowels again? or the last two rows of the consonant tables are approximants, so /r,l,j/ are approximants.

Some features have alternative feature sets in order to accommodate degrees of aggregation. For example, /r,l,j,w/ are approximant in manner but divided into liquid and glide in manner_alt.

Consonants

data_features_consonants is a dataframe with 24 rows and 10 variables.

knitr::kable(data_features_consonants)

phone	cmubet	wiscbet	voicing	manner	manner_alt	place	place_fct	sonorance	sonorance_alt
p	P	p	voiceless	stop	stop	labial	labial	obstruent	obstruent
b	B	b	voiced	stop	stop	labial	labial	obstruent	obstruent
t	T	t	voiceless	stop	stop	alveolar	alveolar	obstruent	obstruent
d	D	d	voiced	stop	stop	alveolar	alveolar	obstruent	obstruent
k	K	k	voiceless	stop	stop	velar	velar	obstruent	obstruent
g	G	g	voiced	stop	stop	velar	velar	obstruent	obstruent
tʃ	CH	tsh	voiceless	affricate	affricate	postalveolar	postalveolar	obstruent	strident
dʒ	JH	dzh	voiced	affricate	affricate	postalveolar	postalveolar	obstruent	strident
m	M	m	voiced	nasal	nasal	labial	labial	sonorant	sonorant
n	N	n	voiced	nasal	nasal	alveolar	alveolar	sonorant	sonorant
ŋ	NG	ng	voiced	nasal	nasal	velar	velar	sonorant	sonorant
f	F	f	voiceless	fricative	fricative	labiodental	labiodental	obstruent	strident
v	V	v	voiced	fricative	fricative	labiodental	labiodental	obstruent	strident
θ	TH	th	voiceless	fricative	fricative	dental	dental	obstruent	obstruent
ð	DH	dh	voiced	fricative	fricative	dental	dental	obstruent	obstruent
s	S	s	voiceless	fricative	fricative	alveolar	alveolar	obstruent	strident
z	Z	z	voiced	fricative	fricative	alveolar	alveolar	obstruent	strident
ʃ	SH	sh	voiceless	fricative	fricative	postalveolar	postalveolar	obstruent	strident
ʒ	ZH	zh	voiced	fricative	fricative	postalveolar	postalveolar	obstruent	strident
h	HH	h	voiceless	fricative	fricative	glottal	glottal	obstruent	obstruent
l	L	l	voiced	approximant	liquid	alveolar	alveolar	sonorant	sonorant
r	R	r	voiced	approximant	liquid	postalveolar	postalveolar	sonorant	sonorant
w	W	w	voiced	approximant	glide	labiovelar	NA	sonorant	sonorant
j	Y	j	voiced	approximant	glide	palatal	palatal	sonorant	sonorant

Description of each column:

phone: phone in IPA
cmubet: phone in the CMU alphabet
wiscbet: phone in an older system used by our lab
voicing: voiced versus voiceless
voicing_alt: spread_glottis versus plain
manner: manner of articulation
manner_alt: alternative manner coding that separates approximants into liquids and glides
place: place of articulation
place_fct: place coded as a factor and ordered based on frontness of the articulators. labiovelar is recoded as NA.
sonorance: obstruent versus sonorant
sonorance_alt: obstruant versus sonorant versus strident.

Levels of the factor columns:

data_features_consonants |>
  lapply(levels) |>
  Filter(length, x = _)
#> $place_fct
#> [1] "labial"       "labiodental"  "dental"       "alveolar"     "postalveolar"
#> [6] "palatal"      "velar"        "glottal"

Considerations about consonant features

The CMU alphabet does not include a glottal stop.

Here /f,v/ are coded as strident following Wikipedia and Sound Pattern of English. If this feature value doesn't seem right, we should probably use an alternative feature of sibilant for the stridents minus /f,v/.

The alternative voicing scheme was suggested by a colleague because of how the voice-voiceless phonetic contrast is achieved with different articulatory strategies in different languages. Note that voicing_alt does not assign a feature to nasals or approximants.

Vowels

data_features_vowels is a dataframe with 17 rows and 11 variables.

knitr::kable(data_features_vowels)

phone	cmubet	wiscbet	hint	manner	manner_alt	tenseness	height	height_fct	backness	backness_fct	rounding
i	IY	i	beat	vowel	vowel	tense	high	high	front	front	unrounded
ɪ	IH	I	bit	vowel	vowel	lax	high	high	front	front	unrounded
eɪ	EY	eI	bait	vowel	vowel	tense	mid	mid	front	front	unrounded
ɛ	EH	E	bet	vowel	vowel	lax	mid	mid	front	front	unrounded
æ	AE	ae	bat	vowel	vowel	lax	low	low	front	front	unrounded
ʌ	AH	^	but	vowel	vowel	lax	mid	mid	central	central	unrounded
ə	AH	4	comma	vowel	vowel	lax	mid	mid	central	central	unrounded
u	UW	u	boot	vowel	vowel	tense	high	high	back	back	rounded
ʊ	UH	U	book	vowel	vowel	lax	high	high	back	back	rounded
oʊ	OW	oU	boat	vowel	vowel	tense	mid	mid	back	back	rounded
ɔ	AO	c	bought	vowel	vowel	tense	mid	mid	back	back	rounded
ɑ	AA	@	bot	vowel	vowel	tense	low	low	back	back	unrounded
aʊ	AW	@U	bout	vowel	diphthong	diphthong	diphthong	NA	diphthong	NA	diphthong
aɪ	AY	@I	bite	vowel	diphthong	diphthong	diphthong	NA	diphthong	NA	diphthong
ɔɪ	OY	cI	boy	vowel	diphthong	diphthong	diphthong	NA	diphthong	NA	diphthong
ɝ	ER	3^	letter	vowel	r-colored	r-colored	mid	mid	central	central	r-colored
ɚ	ER	4^	burt	vowel	r-colored	r-colored	mid	mid	central	central	r-colored

Description of each column:

phone: phone in IPA
cmubet: phone in the CMU alphabet
wiscbet: phone in an older system used by our lab
hint: a word containing the selected vowel
manner: manner of articulation
manner_alt: alternative manner with vowel, diphthong and r-colored
tenseness: tense versus lax (versus diphthong and r-colored)
height: vowel height
height_fct: height coded as a factor ordered high, mid, low. diphthong is recoded to NA.
backness: vowel backness
backness_fct: backness coded as a factor ordered front, central, back. diphthong is recoded to NA.
rounding: unrounded versus rounded (versus diphthong and r-colored)

Levels of the factor columns:

data_features_vowels |>
  lapply(levels) |>
  Filter(length, x = _)
#> $height_fct
#> [1] "high" "mid"  "low"
#>
#> $backness_fct
#> [1] "front"   "central" "back"

Considerations about vowel features

I don't consider /eɪ/ and /oʊ/ to be diphthongs, but perhaps manner_alt could encode the difference of these vowels from the others.

In the CMU alphabet and ARPAbet, vowels can include a number to indicate vowel stress, so AH1 or AH2 is /ʌ/ but AH0 is /ə/.

The vowel features for General American English, according to Wikipedia, are as follows:

	Front		Central		Back
	lax	tense	lax	tense	lax	tense
Close	ɪ	i			ʊ	u
Mid	ɛ	eɪ	ə	(ɜ)		oʊ
Open	æ		(ʌ)	ɑ		(ɔ)
Diphthongs	aɪ ɔɪ aʊ

I adapted these features in this way:

tense and lax features were directly borrowed. Diphthongs and r-colored vowels are were not assign a tenseness.
/ʌ,ɔ/ raised to mid (following the general IPA chart)
/ɑ/ moved to back (following the general IPA)
diphthongs have no backness or height
r-colored vowels were given the backness and height of the /ʌ,ə/

Based on the assumption that /ʌ,ə/ are the same general vowel with differing stress, these vowels have the same features. This definition clashes with the general IPA chart which places /ʌ/ as a back vowel. However, /ʌ/ is a conventional notation. Quoting Wikipedia again: "Although the notation /ʌ/ is used for the vowel of STRUT in RP and General American, the actual pronunciation is closer to a near-open central vowel [ɐ] in RP and advanced back [ʌ̟] in General American." That is, /ʌ/ is fronted in American English (hence, mid) in American English.