English (US) ARPA dictionary v2.0.0a#
@article{gorman2011prosodylab,
author={Gorman, Kyle and Howell, Jonathan and Wagner, Michael},
title={Prosodylab-aligner: A tool for forced alignment of laboratory speech},
journal={Canadian Acoustics},
volume={39},
number={3},
pages={192--193},
year={2011}
}
G2P models Acoustic models |
Installation#
Install from the MFA command line:
mfa model download dictionary english_us_arpa
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 English transcripts.
This dictionary uses the ARPA phone set for English, and was used in training the English ARPA 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: 44,595 Examples: * bhamo: [B A A 1 M O W 0] * meth: [M E H 1 T H] * ahma: [A A 1 M A H 0] * milne: [M I H 1 L N] |
Occurrences: 94,848 Examples: * fanny: [F A E 1 N I Y 0] * r'man: [R M A H 0 N] * nteya: [N T E Y 1 A H 0] * vain: [V E Y 1 N] |
Occurrences: 15,230 Examples: * tunks: [T A H 1 N G K S] * engle: [E H 1 N G G A H 0 L] * zank: [Z A E 1 N G K] * swung: [S W A H 1 N G] |
|||||
Stop |
Occurrences: 32,322 Examples: * akpan: [A E 1 K P A E 2 N] * payne: [P E Y 1 N] * spes: [S P S] * seeps: [S I Y 1 P S] Occurrences: 31,562 Examples: * bajus: [B A E 1 J H A H 0 S] * bepi: [B I Y 1 P I Y 0] * greeb: [G R I Y 1 B] * eban: [E H 1 B A H 0 N] |
Occurrences: 75,542 Examples: * vert: [V E R 1 T] * upsot: [A H 1 P S A H 0 T] * mult: [M A H 2 L T] * tokay: [T O W 0 K E Y 1] Occurrences: 52,334 Examples: * sam'd: [S A E 1 M D] * wode: [W O W 1 D] * beida: [B A Y 1 D A H 0] * glyde: [G L A Y 1 D] |
Occurrences: 58,375 Examples: * carly: [K A A 1 R L I Y 0] * ken's: [K E H 1 N Z] * lawks: [L A O 1 K S] * brocq: [B R A A 1 K] Occurrences: 20,766 Examples: * doing: [D U W 1 I H 0 N G] * logs: [L A O 1 G Z] * suger: [S U W 1 G E R 0] * gift: [G I H 1 F T] |
|||||
Affricate |
Occurrences: 8,420 Examples: * chime: [C H A Y 1 M] * lauch: [L A A 1 C H] * chiz: [C H I H 1 Z] * chers: [C H E R 0 Z] Occurrences: 10,111 Examples: * joly: [J H O W 1 L I Y 0] * jorce: [J H A O 1 R S] * moejy: [M O W 1 J H I Y 0] * jadoo: [J H A H 0 D U W 2] |
|||||||
Sibilant |
Occurrences: 78,674 Examples: * sealy: [S I Y 1 L I Y 0] * pers: [P E R 1 S] * shee: [S H I Y 1] * suas: [S W A A 0 Z] Occurrences: 53,038 Examples: * ising: [A Y 2 Z I H 0 N G] * ize: [A Y 1 Z] * hag's: [H H A E 1 G Z] * banze: [B A E 1 N Z] |
Occurrences: 12,288 Examples: * ships: [S H I H 1 P S] * swash: [S W A A 1 S H] * shaka: [S H A A 1 K A H 0] * thash: [T H A E 1 S H] Occurrences: 796 Examples: * draj: [D R A A 1 Z H] * giron: [Z H I H 1 R A H 0 N] * gime: [Z H I Y 1 M] * zhall: [Z H A H 0 L] |
||||||
Fricative |
Occurrences: 20,957 Examples: * fins: [F I H 1 N Z] * harf: [H H A A 1 R F] * oft: [A O 1 F T] * firin: [F A Y 1 R I H 0 N] Occurrences: 17,526 Examples: * novak: [N O W 1 V A E 0 K] * vegin: [V E H 1 J H I H 0 N] * verts: [V E R 1 T S] * evors: [E H 1 V E R 0 Z] |
Occurrences: 6,592 Examples: * seth: [S E H 1 T H] * tha'd: [T H A O 1 D] * thraw: [T H R A O 1] * thoul: [T H U W 1 L] Occurrences: 1,311 Examples: * those: [D H O W 1 Z] * paths: [P A E 1 D H Z] * lathe: [L E Y 1 D H] * dthe: [D D H] |
Occurrences: 13,675 Examples: * lehua: [L I Y 1 H H Y U W 0 A H 0] * whim: [H H W I H 1 M] * hefty: [H H E H 1 F T I Y 0] * holf: [H H O W 1 L F] |
|||||
Approximant |
Occurrences: 14,822 Examples: * swen: [S W E H 1 N] * gasko: [G A E 1 S K O W 0] * obi: [O W 1 B I Y 0] * hwfa: [H H W F A H 0] |
Occurrences: 71,342 Examples: * tigre: [T I H 0 G R] * rac: [R A E 1 K] * trier: [T R A Y 1 E R 0] * frean: [F R I Y 1 N] |
Occurrences: 8,159 Examples: * niti: [N I H 1 T I Y 0] * reims: [R I Y 1 M Z] * mica: [M A Y 1 K A H 0] * mitzi: [M I H 1 T S I Y 0] |
|||||
Lateral |
Occurrences: 76,788 Examples: * clost: [K L A O 1 S T] * tirl: [T E R 1 L] * decle: [D A H 0 K L] * jumel: [J H U W 1 M A H 0 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: 55,169 Examples: * kauri: [K A O 1 R I Y 0] * wyley: [W A Y 1 L I Y 0] * feete: [F I Y 1 T] * hakim: [A A 0 K I Y 1 M] |
Occurrences: 16,256 Examples: * neuse: [N U W 1 S] * neuk: [N U W 1 K] * duena: [D U W 0 E H 1 N A H 0] * d'eu: [D E Y 2 U W 0] |
|||
Occurrences: 79,698 Examples: * beric: [B E H 1 R I H 0 K] * shear: [S H I H 1 R] * tring: [T R I H 0 N G] * bonis: [B O W 1 N I H 0 S] |
Occurrences: 4,024 Examples: * looka: [L U H 1 K A H 0] * surer: [S H U H 1 R E R 0] * fulls: [F U H 1 L Z] * rawul: [R A E 2 W U H 2 L] |
||||
Close-Mid |
Occurrences: 21,225 Examples: * guere: [G E H 1 R E Y 0] * pay's: [P E Y 1 Z] * nevuh: [N E Y 0 V A H 1] * eira: [E Y 1 R A H 0] |
Occurrences: 25,875 Examples: * aulad: [O W 1 L A E 0 D] * oles: [O W 2 L E Y 1 Z] * ganso: [G A E 1 N S O W 0] * servo: [S E R 1 V O W 0] |
|||
Occurrences: 116,953 Examples: * botha: [B O W 1 T A H 2] * cabul: [K A E 1 B Y A H 0 L] * tums: [T A H 1 M Z] * yum: [Y A H 1 M] |
|||||
Open-Mid |
Occurrences: 39,703 Examples: * elena: [E H 2 L E Y 1 N A H 0] * bizet: [B I H 0 Z E H 1 T] * betha: [B E H 1 T H A H 0] * karat: [K E H 1 R A H 0 T] |
||||
Occurrences: 34,206 Examples: * tabac: [T A H 0 B A E 2 K] * natly: [N A E 1 T L I Y 0] * cram: [K R A E 1 M] * laxer: [L A E 1 K S E R 0] |
Occurrences: 40,624 Examples: * urfa: [Y E R 2 F A H 0] * somer: [S A H 1 M E R 0] * rire: [R A Y 1 E R 0] * berl: [B E R 1 L] |
||||
Open |
Occurrences: 37,763 Examples: * narth: [N A A 2 R T H] * narau: [N A A 2 R A W 1] * ca'y: [K A A 0 I Y 0] * ancha: [A A 0 N K A H 0] Occurrences: 18,014 Examples: * georg: [G E Y 1 A O 0 R G] * snaw: [S N A O 2] * aussi: [A O 1 S I Y 0] * amore: [A A 1 M A O 0 R] |
Diphthongs#
AW
AY
OY
Stress#
0
1
2