eng-syl

# English Syllabifier (eng_syl) This is a GRU-based neural network designed for English word syllabification. The model was trained on data from the Wikimorph dataset.

Usage

Use the syllabify() function from the Syllabel class to syllabify your words:

>>> from eng_syl.syllabify import Syllabel
>>> syllabler = Syllabel()
>>> syllabler.syllabify("chomsky")
'chom-sky'

syllabify() parameters

• text: string- English text to be syllabified. Input should only contain alphabetic characters.

syllabify() returns the given word with hyphens inserted at syllable boundaries.

Onceler (Onset, Nucleus, Coda Segmenter)

The onc_split() function from the Onceler class splits single syllables into their constituent Onset, Nucleus, and Coda components.

>>> from eng_syl.onceler import Onceler
>>> onc = Onceler()
>>> print(onc.onc_split("schmear")
'schm-ea-r'

• text: string - English single syllable word/ component to be segmented into Onset, Nucleus, Coda. Input should only contain alphabetic characters.

Phonify (Grapheme sequence to IPA estimation)

The ipafy() function from the onc_to_phon class tries to approximate an IPA pronunciation from a sequence of graphemes.

>>> from eng_syl.phonify import onc_to_phon
>>> otp = onc_to_phon()
>>> print(otp.ipafy(['schm', 'ea', 'r'])
['ʃm', 'ɪ', 'r']

• sequence: array of strings - sa sequence of English viable onsets, nuclei, and coda

4.0.2 Notes

Fixed a typo in build_model(), where improper shape was being passed into Input() Reverted class name from Syllabel -> Syllable -> Syllabel

4.0.3 Notes

Added handling for non-alpha characters in string; syllabify() won't break immediately if you pass a string like 'he23llotruc38k'. Instead, syllabify() syllabifies the string, ignoring non-alpha characters, and reinserts the non-alpha characters with hyphenation -> 'he23l-lo-truc38k'. This allows for handling of prehyphenated words like 'u-turn' -> 'u--turn'. Also added an arg for returning the syllables as a list in syllabify(word, return_list = False). Should be capable of handling most strings now.

4.0.4 Notes

Added arg save_clean to syllabify(word, save_clean = True). When save_clean, new words will be saved to self.clean for future reference.

4.0.7 Notes

Added evaluate_english_validity(syllable) to Syllabel, which returns the onset nucleus coda decomposition split by hyphens if the syllable is likely English pronounceable, and False if unlikely.

4.0.7 Notes

Fixed casing issues by adding a check to verify only lowercase strings, and consistently reinsert syllable coundaries regardless of original casing.