@readapt/dictionaries

Generate dictionaries

To regenerate dictionaries run scripts (only need if sources in /data changes)

# generate fr dictionary
yarn generate-fr-dict
# updates src/en/en.json file

# generate en dictionary
yarn generate-en-dict
# updates src/fr/fr.json file

Update dictionaries

Updates can be done by adding words in scripts/data/enUpdate.csv for English and scripts/data/frUpdate.csv for French. A new word will be added or overridden if word already exists.

Accepted format is Word,WordAnalysis where Word is the word in lowercase and WordAnalysis is the phonetic transcription where syllables are separated by / and phonemes are separated by .

Example of line for the word malvoyant in French:

malvoyant,m.a.l/v.oi/y.an.an._

Silent letters uses _ phoneme. Special phoneme # is used to signify a non-aspirated h ("h aspiré") where the French liaison is not permitted.

If a letter has more than one phoneme, phonemes are separated by the symbol |

Example of a word with more than one phoneme in a letter

prism,p.r.ih.z|ah/m

Regenerate dictionary and rebuild it running yarn build to take updates account.

Word structure

Each dictionary is a JSON Object which maps each word with their WordAnalysis

A WordAnalysis is an array with 2 or 3 items

• Phoneme index (aligned with graphemes)
• Syllable positions
• Silent letter positions (if needed)

For example the analysis of the word "thoughtful" is :

{
  "thoughtful": [
    // phonemes index aligned with graphemes index
    // th(32) - ao(4) - t(31) - f(14) - ah(3) - l(21)
    // t,  h, o, u, g, h,  t,  f, u, l
    [ 32, 32, 4, 4, 0, 0, 31, 14, 3, 21 ],
    // syllables positions
    [ [ 0, 6 ], [ 7, 9 ] ], // thought/ful
    // silent letter positions
    [ 4, 5 ] // 'gh'
  ]
}