Swahili MFA dictionary v2.0.0#
@techreport{mfa_swahili_mfa_dictionary_2022,
author={McAuliffe, Michael and Sonderegger, Morgan},
title={Swahili MFA dictionary v2.0.0},
address={\url{https://mfa-models.readthedocs.io/pronunciation dictionary/Swahili/Swahili MFA dictionary v2_0_0.html}},
year={2022},
month={Mar},
}
G2P models Acoustic models |
Installation#
Install from the MFA command line:
mfa model download dictionary swahili_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 Swahili transcripts.
This dictionary uses the MFA phone set for Swahili, and was used in training the Swahili 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 |
Glottal |
---|---|---|---|---|---|---|---|---|
Nasal |
Occurrences: 12,949 Examples: * thoma: [t ɔ m ɑ] * mwazo: [m w ɑ z ɔ] * mtito: [m t i t ɔ] * roma: [ɾ ɔ m ɑ] |
Occurrences: 213 Examples: * mfuko: [ɱ f u k ɔ] * mvumi: [ɱ v u m i] * mfike: [ɱ f i k ɛ] * mfu: [ɱ f u] |
Occurrences: 14,525 Examples: * fini: [f i n i] * nati: [n ɑ t i] * keino: [k ɛ i n ɔ] * ndovi: [n d ɔ v i] |
Occurrences: 1,477 Examples: * penye: [p ɛ ɲ ɛ] * nyla: [ɲ l ɑ] * nyayo: [ɲ ɑ j ɔ] * nyaya: [ɲ ɑ j ɑ] |
Occurrences: 3,695 Examples: * nguo: [ŋ ɡ u ɔ] * bunge: [b u ŋ ɡ ɛ] * bingu: [b i ŋ ɡ u] * ngazi: [ŋ ɡ ɑ z i] |
|||
Stop Plain |
Occurrences: 4,340 Examples: * penye: [p ɛ ɲ ɛ] * fupi: [f u p i] * haipo: [h ɑ i p ɔ] * picha: [p i tʃ ɑ] Occurrences: 2,435 Examples: * boas: [b ɔ ɑ s] * jambo: [dʒ ɑ m b ɔ] * bega: [b ɛ ɠ ɑ] * baya: [b ɑ j ɑ] |
Occurrences: 10,702 Examples: * utatu: [u t ɑ t u] * thoma: [t ɔ m ɑ] * mtito: [m t i t ɔ] * fito: [f i t ɔ] Occurrences: 2,643 Examples: * world: [w ɔ ɾ l d] * ndovi: [n d ɔ v i] * pinde: [p i n d ɛ] * dodie: [d ɔ ɗ i ɛ] |
Occurrences: 18,104 Examples: * wako: [w ɑ k ɔ] * kede: [k ɛ ɗ ɛ] * keino: [k ɛ i n ɔ] * ice: [i k ɛ] Occurrences: 3,927 Examples: * nguo: [ŋ ɡ u ɔ] * gairo: [ɡ ɑ i ɾ ɔ] * games: [ɡ ɑ m ɛ s] * gion: [ɡ i ɔ n] |
|||||
Implosive |
Occurrences: 2,675 Examples: * aruba: [ɑ ɾ u ɓ ɑ] * abed: [ɑ ɓ ɛ d] * klabu: [k l ɑ ɓ u] * taabu: [t ɑː ɓ u] |
Occurrences: 1,731 Examples: * rudi: [ɾ u ɗ i] * video: [v i ɗ ɛ ɔ] * kede: [k ɛ ɗ ɛ] * leeds: [l ɛː ɗ s] |
Occurrences: 1,774 Examples: * wigo: [w i ɠ ɔ] * bega: [b ɛ ɠ ɑ] * magma: [m ɑ ɠ m ɑ] * mpige: [m p i ɠ ɛ] |
|||||
Affricate |
Occurrences: 1,998 Examples: * picha: [p i tʃ ɑ] * rich: [ɾ i tʃ] * chapa: [tʃ ɑ p ɑ] * tupac: [t u p ɑ tʃ] Occurrences: 3,450 Examples: * jela: [dʒ ɛ l ɑ] * jaluo: [dʒ ɑ l u ɔ] * mrija: [m ɾ i dʒ ɑ] * jambo: [dʒ ɑ m b ɔ] |
|||||||
Sibilant |
Occurrences: 5,444 Examples: * siasa: [s i ɑ s ɑ] * boas: [b ɔ ɑ s] * ofisi: [ɔ f i s i] * sisi: [s i s i] Occurrences: 6,430 Examples: * mwazo: [m w ɑ z ɔ] * izizi: [i z i z i] * kiza: [k i z ɑ] * zana: [z ɑ n ɑ] |
Occurrences: 4,258 Examples: * flash: [f l ɑ ʃ] * asha: [ɑ ʃ ɑ] * shahi: [ʃ ɑ h i] * ishi: [i ʃ i] |
||||||
Fricative |
Occurrences: 3,364 Examples: * ofisi: [ɔ f i s i] * fini: [f i n i] * fito: [f i t ɔ] * flash: [f l ɑ ʃ] Occurrences: 2,189 Examples: * video: [v i ɗ ɛ ɔ] * ndovi: [n d ɔ v i] * mvumi: [ɱ v u m i] * rave: [ɾ ɑ v ɛ] |
Occurrences: 314 Examples: * south: [s ɑ u θ] * uthai: [u θ ɑ i] * smith: [s m i θ] * nithi: [n i θ i] Occurrences: 574 Examples: * harod: [h ɑ ɾ ɔ ð] * faud: [f ɑ u ð ɑ] * badhi: [b ɑ ð i] * adrid: [ɑ ɗ ɾ i ð ɑ] |
Occurrences: 4,836 Examples: * haipo: [h ɑ i p ɔ] * huwa: [h u w ɑ] * johan: [dʒ ɔ h ɑ n] * hamid: [h ɑ m i d] |
|||||
Approximant |
Occurrences: 13,210 Examples: * mwazo: [m w ɑ z ɔ] * wako: [w ɑ k ɔ] * waoo: [w ɑ ɔː] * world: [w ɔ ɾ l d] |
Occurrences: 3,913 Examples: * yetu: [j ɛ t u] * taya: [t ɑ j ɑ] * kuya: [k u j ɑ] * baya: [b ɑ j ɑ] |
||||||
Tap |
Occurrences: 5,273 Examples: * aruba: [ɑ ɾ u ɓ ɑ] * rudi: [ɾ u ɗ i] * roma: [ɾ ɔ m ɑ] * mari: [m ɑ ɾ i] |
|||||||
Lateral |
Occurrences: 13,852 Examples: * jela: [dʒ ɛ l ɑ] * jaluo: [dʒ ɑ l u ɔ] * flash: [f l ɑ ʃ] * world: [w ɔ ɾ l d] |
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: 43,427 Examples: * siasa: [s i ɑ s ɑ] * ofisi: [ɔ f i s i] * sisi: [s i s i] * fini: [f i n i] Occurrences: 518 Examples: * nabii: [n ɑ ɓ iː] * vii: [v iː] * manii: [m ɑ n iː] * viini: [v iː n i] |
Occurrences: 21,693 Examples: * utatu: [u t ɑ t u] * aruba: [ɑ ɾ u ɓ ɑ] * south: [s ɑ u θ] * rudi: [ɾ u ɗ i] Occurrences: 289 Examples: * huuza: [h uː z ɑ] * kuuma: [k uː m ɑ] * vikuu: [v i k uː] * mguu: [m ɡ uː] |
|||
Close-Mid |
|||||
Open-Mid |
Occurrences: 15,669 Examples: * jela: [dʒ ɛ l ɑ] * penye: [p ɛ ɲ ɛ] * video: [v i ɗ ɛ ɔ] * kede: [k ɛ ɗ ɛ] Occurrences: 171 Examples: * leeds: [l ɛː ɗ s] * meena: [m ɛː n ɑ] * wazee: [w ɑ z ɛː] * njee: [n dʒ ɛː] |
Occurrences: 12,317 Examples: * boas: [b ɔ ɑ s] * ofisi: [ɔ f i s i] * thoma: [t ɔ m ɑ] * mwazo: [m w ɑ z ɔ] Occurrences: 212 Examples: * waoo: [w ɑ ɔː] * vioo: [v i ɔː] * mloo: [m l ɔː] * noor: [n ɔː ɾ] |
|||
Open |
Occurrences: 62,585 Examples: * siasa: [s i ɑ s ɑ] * boas: [b ɔ ɑ s] * utatu: [u t ɑ t u] * thoma: [t ɔ m ɑ] Occurrences: 1,226 Examples: * naa: [n ɑː] * taabu: [t ɑː ɓ u] * maasi: [m ɑː s i] * kupaa: [k u p ɑː] |