syllabification.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

'''
This module attempts to implement syllabification for Greek words.

The consonant clusters are derived from "Greek: A Comprehensive Grammar of the
Modern Language" by D. Holton, et al.
'''

import phonemes
import re

__all__ = [ 'syllabificate', 'get_stressed_syllable', 'get_vowel',
  'get_prevowel_sound' ]


vowelmodifiers = list(u'ˌˈ')

vowels = list(u'ieaouɪə')
palatals = list(u'jçɲʎ')
consonants = list(u'ptkfθxvδɣszlrmn') + palatals

wordinitial = list(consonants) + [
  u'ps', u'ts', u'ks', u'pr', u'tr', u'kr',
  u'pl', u'pn', u'kl', u'kn',
  u'tm',
  u'mn',
  u'fr', u'fl', u'θl', u'θr', u'xl', u'xr', u'vl', u'δr', u'ɣl', u'ɣr',
  u'sp', u'st', u'sk', u'sf', u'sθ', u'sx',
  u'sn', u'zn', u'sl', u'zl',
  u'zb',
  u'ɣn', u'θn', u'xn', u'zm',
  u'ft', u'fk', u'xt',
  u'vδ', u'vɣ', u'ɣδ', u'zv', u'zɣ',
  u'fθ', u'xθ', u'kt', u'pt',
  u'bl', u'br', u'dr', u'gr', u'gl',
  u'δj', u'νj', u'pç'
  u'spl', u'spr', u'skl', u'str', u'skr',
  u'skn',
  u'sfr'
]

wordmedial = list(wordinitial) + [
  u'tn', u'fn', u'vn', u'θm', u'vm',
  u'rt', u'rδ', u'rf', u'rk', u'rx', u'rn', u'rm', u'rɣ', u'lm', u'lp', u'lk',
    u'lt', u'lɣ',
  u'nθ', u'nx', u'mf', u'ns', u'nδ', u'mv', u'nɣ', u'nz',
  u'sxr', u'sxn', u'sθm', u'ptr',
  u'ktr', u'kst', u'kδr', u'ngl', u'mbr', u'ftr', u'fst', u'vɣl' u'rtr',
  u'kstr', u'fstr'
]


def greedy_split(phonemes):
  patterns = list()
  patterns.extend('^' + x for x in sorted(wordinitial, key=len, reverse=True))
  patterns.extend(sorted(wordmedial, key=len, reverse=True))
  patterns.extend('[' + ''.join(vowelmodifiers) + ']*' + v for v in vowels)
  regex = '(' + '|'.join(patterns) + ')'
  
  result = re.split(regex, ''.join(phonemes), flags=re.UNICODE)
  return [ x for x in result if x != '' ]


def is_vowel(syllable):
  return phonemes.strip_modifiers(syllable) in vowels


def is_consonant_cluster(syllable):
  return all(not is_vowel(p) for p in syllable)


def form_syllables(l):
  syllables, temp = [], ''
  for i, group in enumerate(l):
    # If this is consonant cluster, and it's the last group, tack it onto the
    # previous syllable
    if is_consonant_cluster(group) and i == len(l) - 1:
      syllables[-1] += group
      continue
    
    # If this is a consonant cluster, and the next one is a consonant cluster,
    # tack this onto the previous syllable too
    if is_consonant_cluster(group) and is_consonant_cluster(l[i + 1]):
      syllables[-1] += group
      continue
    
    # If this is a consonant cluster, and next up is a vowel, we get tacked
    # onto that vowel
    if is_consonant_cluster(group):
      assert temp == ''
      temp = group
      continue
    
    # If this is a vowel, then we form a new syllable
    if is_vowel(group):
      syllables.append(temp + group)
      temp = ''
  return tuple(syllables)


def syllabificate(phonemes):
  return form_syllables(greedy_split(phonemes))


def get_stressed_syllable(syllables):
  '''
  Returns a negative index for the stressed syllable (i.e., last is -1)
  '''
  for i in xrange(-1, -len(syllables) - 1, -1):
    if u'ˈ' in syllables[i]:
      return i
  raise Exception('stressed syllable not found')


def get_vowel(syllable):
  for c in syllable:
    if is_vowel(c):
      return c
  raise Exception('vowel not found')


def get_prevowel_sound(syllable):
  syllable = phonemes.strip_modifiers(syllable)
  for i, c in enumerate(syllable):
    if is_vowel(c):
      return syllable[:i-1]
  return ''