def compute_vowel_representation(self):
'''
Compute a representation of the lyrics where only vowels are preserved.
'''
self.vow = [] # Lyrics with all but vowels removed
self.vow_idxs = [] # Indices of the vowels in self.text list
self.word_ends = [] # Indices of the last characters of each word
self.words = [] # List of words in the lyrics
self.line_idxs = []
if len(self.language) >= 2 and self.language[:2] == 'en':
self.text_orig = self.text
self.text = ph.get_phonetic_transcription(self.text, output_fname=self.filename+'.ipa')
self.word_ends_orig = []
self.words_orig = []
prev_space_idx = -1 # Index of the previous space char
line_idx = 0 # Line index of the current character
# Go through the lyrics char by char
for i in range(len(self.text)):
self.line_idxs.append(line_idx)
c = self.text[i]
c = ph.map_vow(c, self.language)
if ph.is_vow(c, self.language):
# Ignore double vowels
# (in English this applies probably only to 'aa' as in 'bath'
# which rhymes with 'trap' that has only 'a')
if i > 0 and self.text[i-1] == c:
# Index of a double vowel points to the latter occurrence
self.vow_idxs[-1] = i
continue
# TODO Diftongs should not be split (i.e. "price" should
# not rhyme with "trap kit"). This has been fixed in BattleBot
self.vow.append(c)
self.vow_idxs.append(i)
elif ph.is_space(c):
if c in '\n':
line_idx += 1
elif c in '.!?' and i < len(self.text)-1 and self.text[i+1] != '\n':
line_idx += 1
# If previous char was not a space, we've encountered word end
if len(self.vow) > 0 and not ph.is_space(self.text[i-1]):
# Put together the new word. Potential consonants in the
# end are ignored
new_word = self.text[prev_space_idx+1:self.vow_idxs[-1]+1]
# Check that the new word contains at least one vowel
no_vowels = True
for c2 in new_word:
if ph.is_vow(c2, self.language):
no_vowels = False
break
if no_vowels:
prev_space_idx = i
continue
self.word_ends.append(len(self.vow)-1)
self.words.append(new_word)
prev_space_idx = i
if len(self.language) >= 2 and self.language[:2] == 'en':
self.lines_orig = self.text_orig.split('\n')
def compute_vowel_representation(self):
'''
Compute a representation of the lyrics where only vowels are preserved.
'''
self.vow = [] # Lyrics with all but vowels removed
self.vow_idxs = [] # Indices of the vowels in self.text list
self.word_ends = [] # Indices of the last characters of each word
self.words = [] # List of words in the lyrics
self.line_idxs = []
if len(self.language) >= 2 and self.language[:2] == 'en':
self.text_orig = self.text
self.text = ph.get_phonetic_transcription(self.text, output_fname=self.filename+'.ipa')
self.word_ends_orig = []
self.words_orig = []
prev_space_idx = -1 # Index of the previous space char
line_idx = 0 # Line index of the current character
# Go through the lyrics char by char
for i in range(len(self.text)):
self.line_idxs.append(line_idx)
c = self.text[i]
c = ph.map_vow(c, self.language)
if ph.is_vow(c, self.language):
# Ignore double vowels
# (in English this applies probably only to 'aa' as in 'bath'
# which rhymes with 'trap' that has only 'a')
if i > 0 and self.text[i-1] == c:
# Index of a double vowel points to the latter occurrence
self.vow_idxs[-1] = i
continue
# TODO Diftongs should not be split (i.e. "price" should
# not rhyme with "trap kit"). This has been fixed in BattleBot
self.vow.append(c)
self.vow_idxs.append(i)
elif ph.is_space(c):
if c in '\n':
line_idx += 1
elif c in '.!?' and i < len(self.text)-1 and self.text[i+1] != '\n':
line_idx += 1
# If previous char was not a space, we've encountered word end
if len(self.vow) > 0 and not ph.is_space(self.text[i-1]):
# Put together the new word. Potential consonants in the
# end are ignored
new_word = self.text[prev_space_idx+1:self.vow_idxs[-1]+1]
# Check that the new word contains at least one vowel
no_vowels = True
for c2 in new_word:
if ph.is_vow(c2, self.language):
no_vowels = False
break
if no_vowels:
prev_space_idx = i
continue
self.word_ends.append(len(self.vow)-1)
self.words.append(new_word)
prev_space_idx = i
if len(self.language) >= 2 and self.language[:2] == 'en':
self.lines_orig = self.text_orig.split('\n')
def rhyme_length_fixed(self, wpos1, wpos2):
'''
Length of rhyme (in vowels). The first part of the rhyme ends with
self.words[wpos1] and the latter part with word self.words[wpos2].
Input:
wpos1 Word index of the last word in the first part of the rhyme.
wpos2 Word index of the end of the rhyme.
'''
if wpos1 < 0: # Don't wrap
return 0
elif self.words[wpos1] == self.words[wpos2]:
return 0
# Indices in the vowel list
p1 = self.word_ends[wpos1]
p2 = self.word_ends[wpos2]
l = 0
while self.vow[p1 - l] == self.vow[p2 - l]:
# Make sure that exactly same words are not used
if wpos1 > 0 and p1 - l <= self.word_ends[
wpos1 -
1] and wpos2 > 0 and p2 - l <= self.word_ends[wpos2 - 1]:
# Get the first and last character indices of the words surrounding the vowels at p1-l and p2-l
prev_s1 = self.vow_idxs[p1 - l]
while prev_s1 > 0 and not ph.is_space(self.text[prev_s1 - 1]):
prev_s1 -= 1
prev_s2 = self.vow_idxs[p2 - l]
while prev_s2 > 0 and not ph.is_space(self.text[prev_s2 - 1]):
prev_s2 -= 1
next_s1 = self.vow_idxs[p1 - l]
while next_s1 < len(self.text) - 1 and not ph.is_space(
self.text[next_s1 + 1]):
next_s1 += 1
next_s2 = self.vow_idxs[p2 - l]
while next_s2 < len(self.text) - 1 and not ph.is_space(
self.text[next_s2 + 1]):
next_s2 += 1
if next_s1 - prev_s1 == next_s2 - prev_s2 and self.text[
prev_s1:next_s1 + 1] == self.text[prev_s2:next_s2 + 1]:
break
l += 1
if p1 - l < 0 or p2 - l <= p1:
break
# Ignore rhymes with length 1
if l == 1:
l = 0
return l
def get_rhyme_str(self, rhyme_tuple):
'''
Construct a string of a given rhyme tuple.
'''
ret = ''
rl, wpos1, wpos2 = rhyme_tuple
if wpos1 is None or wpos2 is None:
return ''
p2 = self.vow_idxs[self.word_ends[wpos2]]
p2_orig = p2
# Find the ending of the last word
while not ph.is_space(self.text[p2]):
p2 += 1
p0 = self.vow_idxs[self.word_ends[wpos1] - rl]
p0_orig = p0
# Find the beginning of the line
while self.text[p0] != '\n' and p0 > 0:
p0 -= 1
cap_line = ''
rw1, rw2 = self.get_rhyming_vowels(rhyme_tuple)
for i in range(p0, p2 + 1):
if i == min(rw1) or i == min(rw2):
cap_line += ' | ' + self.text[i]
elif i == max(rw1) or i == max(rw2):
cap_line += self.text[i] + '|'
else:
cap_line += self.text[i]
ret += "Longest rhyme (l=%d): %s\n" % (rl, cap_line)
return ret
def rhyme_length_fixed(self, wpos1, wpos2):
'''
Length of rhyme (in vowels). The first part of the rhyme ends with
self.words[wpos1] and the latter part with word self.words[wpos2].
Input:
wpos1 Word index of the last word in the first part of the rhyme.
wpos2 Word index of the end of the rhyme.
'''
if wpos1 < 0: # Don't wrap
return 0
elif self.words[wpos1] == self.words[wpos2]:
return 0
# Indices in the vowel list
p1 = self.word_ends[wpos1]
p2 = self.word_ends[wpos2]
l = 0
while self.vow[p1-l] == self.vow[p2-l]:
# Make sure that exactly same words are not used
if wpos1 > 0 and p1-l <= self.word_ends[wpos1-1] and wpos2 > 0 and p2-l <= self.word_ends[wpos2-1]:
# Get the first and last character indices of the words surrounding the vowels at p1-l and p2-l
prev_s1 = self.vow_idxs[p1-l]
while prev_s1 > 0 and not ph.is_space(self.text[prev_s1-1]):
prev_s1 -= 1
prev_s2 = self.vow_idxs[p2-l]
while prev_s2 > 0 and not ph.is_space(self.text[prev_s2-1]):
prev_s2 -= 1
next_s1 = self.vow_idxs[p1-l]
while next_s1 < len(self.text)-1 and not ph.is_space(self.text[next_s1+1]):
next_s1 += 1
next_s2 = self.vow_idxs[p2-l]
while next_s2 < len(self.text)-1 and not ph.is_space(self.text[next_s2+1]):
next_s2 += 1
if next_s1-prev_s1 == next_s2-prev_s2 and self.text[prev_s1:next_s1+1] == self.text[prev_s2:next_s2+1]:
break
l += 1
if p1-l < 0 or p2-l <= p1:
break
# Ignore rhymes with length 1
if l == 1:
l = 0
return l
def find_rhymes(self, vowel):
rhyming_vowel_idxs = [
self.vow_idxs[i] for i in range(len(self.vow_idxs))
if self.vow[i] == vowel
]
rhyming_words = []
rhyming_word_indices = []
for vow_idx in rhyming_vowel_idxs:
start = end = vow_idx
while not ph.is_space(self.text[start]):
start -= 1
while not ph.is_space(self.text[end]):
end += 1
word = self.text[start:end + 1].strip()
#rhyming_word_indices.append(self.espeak_words.index(word))
rhyming_words.append(
self.words_orig[self.espeak_words.index(word)])
#print(rhyming_word_indices)
#rhyming_word = [word for i, word in enumerate(self.words_orig) if i in rhyming_word_indices]
#print(rhyming_vowel_idxs)
return rhyming_words
def get_rhyme_str(self, rhyme_tuple):
'''
Construct a string of a given rhyme tuple.
'''
ret = ''
rl, wpos1, wpos2 = rhyme_tuple
if wpos1 is None or wpos2 is None:
return ''
p2 = self.vow_idxs[self.word_ends[wpos2]]
p2_orig = p2
# Find the ending of the last word
while not ph.is_space(self.text[p2]):
p2 += 1
p0 = self.vow_idxs[self.word_ends[wpos1]-rl]
p0_orig = p0
# Find the beginning of the line
while self.text[p0] != '\n' and p0 > 0:
p0 -= 1
cap_line = ''
rw1, rw2 = self.get_rhyming_vowels(rhyme_tuple)
for i in range(p0,p2+1):
if self.language == 'fi':
if i in rw1 or i in rw2:
cap_line += self.text[i].capitalize()
else:
cap_line += self.text[i]
else:
if i == min(rw1) or i == min(rw2):
cap_line += ' | ' + self.text[i]
elif i == max(rw1) or i == max(rw2):
cap_line += self.text[i] + '|'
else:
cap_line += self.text[i]
#ret += "Longest rhyme (l=%d): %s\n" % (rl, cap_line)
ret += "Longest rhyme (l={0}): {1}\n".format(int(rl), cap_line)
if self.language != 'fi':
# Get the corresponding lines from the original lyrics
line_beg = self.line_idxs[p0]
line_end = self.line_idxs[p2]
for i in range(line_beg, line_end+1):
if i < len(self.lines_orig):
ret += self.lines_orig[i] + '\n'
return ret
def get_rhyme_str(self, rhyme_tuple):
'''
Construct a string of a given rhyme tuple.
'''
ret = ''
rl, wpos1, wpos2 = rhyme_tuple
if wpos1 is None or wpos2 is None:
return ''
p2 = self.vow_idxs[self.word_ends[wpos2]]
p2_orig = p2
# Find the ending of the last word
while not ph.is_space(self.text[p2]):
p2 += 1
p0 = self.vow_idxs[self.word_ends[wpos1]-rl]
p0_orig = p0
# Find the beginning of the line
while self.text[p0] != '\n' and p0 > 0:
p0 -= 1
cap_line = ''
rw1, rw2 = self.get_rhyming_vowels(rhyme_tuple)
for i in range(p0,p2+1):
if self.language == 'fi':
if i in rw1 or i in rw2:
cap_line += self.text[i].capitalize()
else:
cap_line += self.text[i]
else:
if i == min(rw1) or i == min(rw2):
cap_line += ' | ' + self.text[i]
elif i == max(rw1) or i == max(rw2):
cap_line += self.text[i] + '|'
else:
cap_line += self.text[i]
ret += "Longest rhyme (l=%d): %s\n" % (rl, cap_line)
if self.language != 'fi':
# Get the corresponding lines from the original lyrics
line_beg = self.line_idxs[p0]
line_end = self.line_idxs[p2]
for i in range(line_beg, line_end+1):
if i < len(self.lines_orig):
ret += self.lines_orig[i] + '\n'
return ret