def read_elp(self):
self.tree = PrefixTree()
self.pronunciations = {}
words = list(csv.DictReader(open(paths.elp)))
for i, x in enumerate(words):
pron = self.elp2cmu.translate(x['Pron'])
if x[self.frequency_field] not in ['0', 'NULL']:
self.tree.insert(tuple(pron + ['#']),
float(x[self.frequency_field]), x['Word'])
self.pronunciations[x['Word']] = pron
self.tree.calculate_probs()
class SegmentSurprisalTree(object):
frequency_field = 'SUBTLWF'
def __init__(self):
self.elp2cmu = ELP2CMU()
def read_elp(self):
self.tree = PrefixTree()
self.pronunciations = {}
words = list(csv.DictReader(open(paths.elp)))
for i, x in enumerate(words):
pron = self.elp2cmu.translate(x['Pron'])
if x[self.frequency_field] not in ['0', 'NULL']:
self.tree.insert(tuple(pron + ['#']),
float(x[self.frequency_field]), x['Word'])
self.pronunciations[x['Word']] = pron
self.tree.calculate_probs()
def surprisals(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_surprisals(pron)
def entropies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_entropies(pron)
def probabilities(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_probabilities(pron)
def frequencies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_frequencies(pron)
def node_frequencies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_frequencies(pron)
def read_elp(self):
self.tree = PrefixTree()
self.pronunciations = {}
words = list(csv.DictReader(open(paths.elp)))
for i, x in enumerate(words):
pron = self.elp2cmu.translate(x['Pron'])
if x[self.frequency_field] not in ['0', 'NULL']:
self.tree.insert(tuple(pron + ['#']),
float(x[self.frequency_field]), x['Word'])
self.pronunciations[x['Word']] = pron
self.tree.calculate_probs()
class SegmentSurprisalTree(object):
frequency_field = 'SUBTLWF'
def __init__(self):
self.elp2cmu = ELP2CMU()
def read_elp(self):
self.tree = PrefixTree()
self.pronunciations = {}
words = list(csv.DictReader(open(paths.elp)))
for i, x in enumerate(words):
pron = self.elp2cmu.translate(x['Pron'])
if x[self.frequency_field] not in ['0', 'NULL']:
self.tree.insert(tuple(pron + ['#']),
float(x[self.frequency_field]), x['Word'])
self.pronunciations[x['Word']] = pron
self.tree.calculate_probs()
def surprisals(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_surprisals(pron)
def entropies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_entropies(pron)
def probabilities(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_probabilities(pron)
def frequencies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_frequencies(pron)
def node_frequencies(self, word, with_end=True):
pron = self.pronunciations[word]
pron = tuple(pron + ['#'] if with_end else [])
return self.tree.prefix_frequencies(pron)
def get_word_continuations(self, word):
# get pronunciation of the word
try:
word_pron = self.pronunciations[word]
except:
raise NotImplementedError("Word '%s' not found in corpus." % word)
# convert into expected tuple format
word_tuple = tuple(word_pron + ['#'] if True else [])
# extract prefix frequency for target word
continuation_dict = {} # empty dict to populate w/ frequencies
phoneme_count = 0
for phoneme_iter, prefix in self.tree.get_prefixes(word_tuple):
# get frequency of all phoneme continuations
cont_freqs = [x[1] for x in self.tree.get_continuations(prefix)]
# add to dictionary
continuation_dict.update(
{'%s_%s' % (phoneme_count, phoneme_iter): cont_freqs})
# update phoneme count
phoneme_count = phoneme_count + 1
return continuation_dict
def get_uniqueness_point(self, word):
# get word continuation dict
continuation_dict = self.get_word_continuations(word)
# sort key entries
phoneme_list = continuation_dict.keys()
phoneme_list.sort()
# loop through each phoneme entry
for n, phoneme in enumerate(phoneme_list):
conts = continuation_dict[phoneme]
# when there is only one continuation, we have found the UP.
if len(conts) == 1:
return n, phoneme
# if no UP found, return error.
raise ValueError("No uniqueness point found.")
