- Step 0: to randomly generate a grammar
- Step 1: to generate an artificial corpus using the grammar
- Step 2: to extract/calculate a grammar from the artificial corpus
- Step 3: to compare the calculated grammar and the original grammar
- Assumption: Restrictions are only length 2, and local.
# Problem: the length of corpus could affect the accuracy of the calculated restrctions; How to statistically calculate the confidence?
# To do: - Evaluate a word's well-formedness under the calculated restrictions. - Generalize the functions to extend the range of applicable grammars (k = n, non-local/distant restrictions) - auto-correction/completion of syllables, words
- # input: syllable/word output: a classified language
- transducer
# output: acceptance rating of a certain syllable/word - compare the output accuracy of feature&rule-based learner and purely frequency-based learner
phonology models and relevant interfaces
- Free software: 3-clause BSD license
- Documentation: (COMING SOON!) https://yangliustonybrook.github.io/phomo.
- TODO