English (US) ARPA dictionary v3.0.0#
@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}
}
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](https://raw.githubusercontent.com/MontrealCorpusTools/mfa-models/main/dictionary/english/mfa/English (US) ARPA dictionary v3_0_0.dict).
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: 44,929 Examples: * sodom: [S AA1 D AH0 M] * misha: [M IH1 SH AH0] * wombs: [W AA1 M Z] * lamby: [L AE1 M B IY0] |
Occurrences: 95,743 Examples: * lynch: [L IH1 N CH] * mon: [M AA1 N] * omens: [OW1 M AH0 N Z] * wayne: [W EY1 N] |
Occurrences: 15,307 Examples: * gong: [G AO1 NG] * kung: [K AH1 NG] * spank: [S P AE1 NG K] * dung: [D AH1 NG] |
|||||
Stop |
Occurrences: 32,646 Examples: * unpin: [AH0 N P AE1 N] * pests: [P EH1 S T S] * gripe: [G R AY1 P] * peril: [P EH1 R AH0 L] Occurrences: 31,830 Examples: * bolus: [B OW1 L AH0 S] * debut: [D EY0 B Y UW1] * verba: [V EH1 R B AH0] * baum: [B AW1 M] |
Occurrences: 76,538 Examples: * feete: [F IY1 T] * pact: [P AE1 K T] * it': [IH1 T AH0] * blast: [B L AE1 S T] Occurrences: 52,829 Examples: * docs: [D AA1 K S] * fated: [F EY1 T IH0 D] * reddy: [R EH1 D IY0] * cody: [K OW1 D IY0] |
Occurrences: 58,935 Examples: * quilt: [K W IH1 L T] * crean: [K R IY1 N] * ankle: [AE1 NG K AH0 L] * cups: [K AH1 P S] Occurrences: 20,940 Examples: * greg: [G R EH1 G] * galls: [G AO1 L Z] * gal: [G AE1 L] * bogs: [B AA1 G Z] |
|||||
Affricate |
Occurrences: 8,472 Examples: * catch: [K AE1 CH] * beach: [B IY1 CH] * chee: [CH IY1] * bench: [B EH1 N CH] Occurrences: 10,171 Examples: * lige: [L AY1 JH] * bulgy: [B AH1 L JH IY0] * jules: [JH UW1 L Z] * jutes: [JH UW1 T S] |
|||||||
Sibilant |
Occurrences: 79,492 Examples: * ibsen: [IH1 B S AH0 N] * syro: [S AY1 R OW0] * skye: [S K AY1] * vox: [V AA1 K S] Occurrences: 53,319 Examples: * pepys: [P EH1 P IY0 Z] * ogres: [OW1 G ER0 Z] * tease: [T IY1 Z] * leos: [L IY1 OW0 Z] |
Occurrences: 12,390 Examples: * shuns: [SH AH1 N Z] * shred: [SH R EH1 D] * shone: [SH OW1 N] * short: [SH AO1 R T] Occurrences: 798 Examples: * rouge: [R UW1 ZH] * dijon: [D IY0 ZH OW1 N] * asia: [EY1 ZH AH0] * usual: [Y UW1 ZH UW0 AH0 L] |
||||||
Fricative |
Occurrences: 21,273 Examples: * fishy: [F IH1 SH IY0] * faint: [F EY1 N T] * farn: [F AA1 R N] * flaws: [F L AO1 Z] Occurrences: 17,654 Examples: * voted: [V OW1 T AH0 D] * vince: [V IH1 N S] * avast: [AA0 V AA0 S T] * verds: [V ER1 D Z] |
Occurrences: 6,636 Examples: * thole: [TH OW1 L] * goth: [G AA1 TH] * thees: [TH IY1 Z] * ninth: [N AY1 N TH] Occurrences: 1,328 Examples: * thee: [DH IY1] * tha: [DH AH0] * these: [DH IY1 Z] * that: [DH AE1 T] |
Occurrences: 13,794 Examples: * sahib: [S AH0 HH IH1 B] * hagar: [HH EY1 G AA0 R] * hegel: [HH EH1 G AH0 L] * heads: [HH EH1 D Z] |
|||||
Approximant |
Occurrences: 15,035 Examples: * wynn: [W IH1 N] * we'll: [W IY1 L] * weber: [W EH1 B ER0] * quiet: [K W AY1 AH0] |
Occurrences: 71,955 Examples: * eery: [IH1 R IY0] * yuruk: [Y UH1 R IH0 K] * farms: [F AA1 R M Z] * maure: [M AO1 R] |
Occurrences: 8,242 Examples: * yella: [Y EH1 L AH0] * yuki: [Y UW1 K IY0] * fuses: [F Y UW1 Z IH0 Z] * hugi: [HH Y UW1 G IY0] |
|||||
Lateral |
Occurrences: 77,647 Examples: * graal: [G R AA1 L] * lab: [L AE1 B] * allee: [AH0 L IY1] * oral: [AO1 R AH0 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: 55,493 Examples: * willy: [W IH1 L IY0] * leeks: [L IY1 K S] * mie's: [M IY0 Z] * quasi: [K W AA1 S IY0] |
Occurrences: 16,342 Examples: * to't: [T UW0 T] * ute: [Y UW1 T] * danu: [D AA1 N UW0] * chiku: [CH IY1 K UW2] |
|||
Occurrences: 80,466 Examples: * arnim: [AA1 R N IH2 M] * orbis: [AO1 R B IH0 S] * nitro: [N IH1 T R OW0] * indra: [IH2 N D R AH0] |
Occurrences: 4,053 Examples: * good': [G UH1 D] * tush: [T UH1 SH] * would: [W UH1 D] * curae: [K Y UH0 R EY1] |
||||
Close-Mid |
Occurrences: 21,451 Examples: * dumay: [D AH0 M EY0] * jabe: [JH AA1 B EY0] * melee: [M EY1 L EY2] * padre: [P AE1 D R EY2] |
Occurrences: 25,982 Examples: * hondo: [HH AA1 N D OW0] * sligo: [S L AY1 G OW2] * gueux: [G OW1] * tiago: [T IY0 AA1 G OW0] |
|||
Occurrences: 118,104 Examples: * juxta: [JH AH2 K S T AH0] * bason: [B AE1 S AH0 N] * dully: [D AH1 L IY0] * lyon: [L AY1 AH0 N] |
|||||
Open-Mid |
Occurrences: 39,991 Examples: * harim: [HH EH0 R IY1 M] * wor: [D AH1 B EH0 L Y UW1 OW1 AA1 R] * toque: [T AA2 K W EH0] * pepin: [P EH1 P IH0 N] |
||||
Occurrences: 34,399 Examples: * boaz: [B OW1 AE0 Z] * altai: [AE0 L T AY1] * kodak: [K OW1 D AE2 K] * kat's: [K AE2 T S] |
Occurrences: 41,063 Examples: * curds: [K ER1 D Z] * serle: [S ER1 AH0 L] * farer: [F EH2 R ER0] * rigor: [R IH1 G ER0] |
||||
Open |
Occurrences: 38,981 Examples: * ameer: [AA2 M IH1 R] * azara: [AA0 Z AA1 R AH0] * tom's: [T AA1 M Z] * aesop: [IY1 S AA2 P] Occurrences: 18,102 Examples: * kiosk: [K IY1 AO2 S K] * wharf: [HH W AO1 R F] * henri: [AO2 R IY1] * utah: [Y UW1 T AO2] |
Diphthongs#
AW
AY
OY
Stress#
0
1
2