Spanish (Latin America) MFA dictionary v2.0.0#

@techreport{mfa_spanish_latin_america_mfa_dictionary_2022,
	author={McAuliffe, Michael and Sonderegger, Morgan},
	title={Spanish (Latin America) MFA dictionary v2.0.0},
	address={\url{https://mfa-models.readthedocs.io/pronunciation dictionary/Spanish/Spanish (Latin America) MFA dictionary v2_0_0.html}},
	year={2022},
	month={Mar},
}
../../_images/full_logo_yellow.svg

Installation#

Install from the MFA command line:

mfa model download dictionary spanish_latin_america_mfa

Or download from the release page.

The dictionary available from the release page and command line installation has pronunciation and silence probabilities estimated as part acoustic model training (see Silence probability format and training pronunciation probabilities for more information. If you would like to use the version of this dictionary without probabilities, please see the plain dictionary.

Intended use#

This dictionary is intended for forced alignment of Spanish transcripts.

This dictionary uses the MFA phone set for Spanish, and was used in training the Spanish MFA acoustic model. Pronunciations can be added on top of the dictionary, as long as no additional phones are introduced.

Performance Factors#

When trying to get better alignment accuracy, adding pronunciations is generally helpful, especially for different styles and dialects. The most impactful improvements will generally be seen when adding reduced variants that involve deleting segments/syllables common in spontaneous speech. Alignment must include all phones specified in the pronunciation of a word, and each phone has a minimum duration (by default 10ms). If a speaker pronounces a multisyllabic word with just a single syllable, it can be hard for MFA to fit all the segments in, so it will lead to alignment errors on adjacent words as well.

Ethical considerations#

Deploying any Speech-to-Text model into any production setting has ethical implications. You should consider these implications before use.

Demographic Bias#

You should assume every machine learning model has demographic bias unless proven otherwise. For pronunciation dictionaries, it is often the case that transcription accuracy and lexicon coverage for the prestige variety modeled in this dictionary compared to other variants. If you are using this dictionary in production, you should acknowledge this as a potential issue.

IPA Charts#

Consonants#

Obstruent symbols to the left of are unvoiced and those to the right are voiced.

Manner

Labial

Labiodental

Dental

Alveolar

Alveopalatal

Palatal

Velar

Nasal

Occurrences:
19,291
Examples:
* muña:
[m u ɲ a]
* agamí:
[a ɣ a m i]
* amaño:
[a m a ɲ o]
* cómic:
[k o m i k]
Occurrences:
30,490
Examples:
* lunes:
[l u n e s]
* pindo:
[p i n o]
* nepal:
[n e p a l]
* anida:
[a n i ð a]
Occurrences:
2,216
Examples:
* muña:
[m u ɲ a]
* niega:
[ɲ j e ɣ a]
* amaño:
[a m a ɲ o]
* chuño:
[ u ɲ o]
Occurrences:
2,304
Examples:
* fango:
[f a ŋ ɡ o]
* monk:
[m o ŋ k]
* zinc:
[s i ŋ k]
* lonco:
[l o ŋ k o]

Stop

Occurrences:
14,201
Examples:
* paula:
[p a u l a]
* pindo:
[p i n o]
* nepal:
[n e p a l]
* papúa:
[p a p u a]
Occurrences:
5,810
Examples:
* vic:
[b i k]
* bromo:
[b ɾ o m o]
* brotó:
[b ɾ o o]
* baza:
[b a s a]
Occurrences:
30,654
Examples:
* nutra:
[n u ɾ a]
* turín:
[ u ɾ i n]
* altea:
[a l e a]
* brotó:
[b ɾ o o]
Occurrences:
8,817
Examples:
* pindo:
[p i n o]
* danta:
[ a n a]
* donna:
[ o n a]
* delta:
[ e l a]
Occurrences:
2,048
Examples:
* kyle:
[c i l]
* kits:
[c i s]
* kent:
[c e n ]
* okey:
[o c e i]
Occurrences:
165
Examples:
* guía:
[ɟ i a]
* guiso:
[ɟ i s o]
* guías:
[ɟ i a s]
* gay:
[ɟ e i]
Occurrences:
20,171
Examples:
* vic:
[b i k]
* case:
[k a s e]
* sioux:
[s j u k s]
* cómic:
[k o m i k]
Occurrences:
2,136
Examples:
* gorra:
[ɡ o r a]
* fango:
[f a ŋ ɡ o]
* good:
[ɡ u ð]
* gonna:
[ɡ o n a]

Affricate

Occurrences:
2,409
Examples:
* chuy:
[ u i]
* cuchí:
[k u i]
* chuño:
[ u ɲ o]
* loach:
[l o ]
Occurrences:
495
Examples:
* yarda:
[ɟʝ a r ð a]
* gyula:
[ɟʝ u l a]
* lleva:
[ɟʝ e β a]
* llamé:
[ɟʝ a m e]

Sibilant

Occurrences:
51,353
Examples:
* salga:
[s a l ɣ a]
* lunes:
[l u n e s]
* azada:
[a s a ð a]
* sisé:
[s i s e]
Occurrences:
263
Examples:
* crash:
[k ɾ a ʃ]
* zeus:
[ ʃ e u s]
* zero:
[ ʃ e ɾ o]
* hash:
[a ʃ]

Fricative

Occurrences:
6,558
Examples:
* fénix:
[f e n i k s]
* focus:
[f o k u s]
* fango:
[f a ŋ ɡ o]
* afile:
[a f i l e]
Occurrences:
26,919
Examples:
* azada:
[a s a ð a]
* anida:
[a n i ð a]
* yarda:
[ɟʝ a r ð a]
* pudo:
[p u ð o]
Occurrences:
2,798
Examples:
* genil:
[ç e n i l]
* gill:
[ç i l]
* orgía:
[o r ç i a]
* aloje:
[a l o ç e]
Occurrences:
1,828
Examples:
* hayas:
[a ʝ a s]
* praia:
[p ɾ a ʝ a]
* poyo:
[p o ʝ o]
* troya:
[ ɾ o ʝ a]

Approximant

Occurrences:
3,103
Examples:
* cuán:
[k w a n]
* suino:
[s w i n o]
* actuó:
[a ɣ w o]
* juega:
[x w e ɣ a]
Occurrences:
9,057
Examples:
* niega:
[ɲ j e ɣ a]
* sioux:
[s j u k s]
* siega:
[s j e ɣ a]
* indie:
[i n j e]

Tap

Occurrences:
31,646
Examples:
* nutra:
[n u ɾ a]
* turín:
[ u ɾ i n]
* bromo:
[b ɾ o m o]
* brotó:
[b ɾ o o]

Trill

Occurrences:
14,859
Examples:
* gorra:
[ɡ o r a]
* yarda:
[ɟʝ a r ð a]
* sorgo:
[s o r ɣ o]
* reían:
[r e i a n]

Lateral

Occurrences:
20,492
Examples:
* salga:
[s a l ɣ a]
* lunes:
[l u n e s]
* paula:
[p a u l a]
* nepal:
[n e p a l]

Vowels#

Vowel symbols to the left of are unrounded and those to the right are rounded.

Front

Near-Front

Central

Near-Back

Back

Close

Occurrences:
37,932
Examples:
* pindo:
[p i n o]
* vic:
[b i k]
* anida:
[a n i ð a]
* sisé:
[s i s e]
Occurrences:
13,415
Examples:
* muña:
[m u ɲ a]
* lunes:
[l u n e s]
* paula:
[p a u l a]
* nutra:
[n u ɾ a]

Close-Mid

Occurrences:
54,952
Examples:
* lunes:
[l u n e s]
* nepal:
[n e p a l]
* sisé:
[s i s e]
* niega:
[ɲ j e ɣ a]
Occurrences:
55,389
Examples:
* pindo:
[p i n o]
* gorra:
[ɡ o r a]
* amaño:
[a m a ɲ o]
* cómic:
[k o m i k]

Open-Mid

Open

Occurrences:
93,500
Examples:
* muña:
[m u ɲ a]
* salga:
[s a l ɣ a]
* paula:
[p a u l a]
* azada:
[a s a ð a]