def Psubs(clsdict, corpus, minlength=3, maxlength=20):
"""
Generate Psubs - the substring completion of a set of pairs.
Psubs is the association between all substrings of the pairs in classdict.
Parameters
----------
clsdict : ClassDict
corpus : Corpus
minlength : int, optional
minimum number of phones for the substrings
maxlength : int, optional
maximum number of phones for the substrings
Returns
-------
Iterator over (FragmentToken, FragmentToken) pairs
"""
sub_pairs = (pairwise_substring_completion(f1, f2, corpus,
minlength,
maxlength)
for f1, f2 in clsdict.iter_pairs(within=True, order=True))
return unique(flatten(sub_pairs))
def iter_fragments(self, with_class=False):
"""
Iterate over FragmentTokens.
Parameters
----------
with_class : bool, optional
Iterate over (ClassID, FragmentToken) pairs instead
Returns
-------
Iterator over FragmentToken or (ClassID, FragmentToken) pairs
"""
if with_class:
return unique(flatten(izip(repeat(c), v)
for c, v in self.clsdict.iteritems()))
else:
return unique(flatten(self.clsdict.itervalues()))
def iter_fragments(self, with_class=False):
"""
Iterate over FragmentTokens.
Parameters
----------
with_class : bool, optional
Iterate over (ClassID, FragmentToken) pairs instead
Returns
-------
Iterator over FragmentToken or (ClassID, FragmentToken) pairs
"""
if with_class:
return unique(
flatten(
izip(repeat(c), v) for c, v in self.clsdict.iteritems()))
else:
return unique(flatten(self.clsdict.itervalues()))
def iter_pairs(self, within, order):
"""
Iterate over FragmentToken pairs.
Parameters
----------
within : bool
Only select pairs from the same class.
order : bool
Also include reverse of a pair.
Returns
-------
Iterator over (FragmentToken, FragmentToken) pairs.
"""
vals = self.clsdict.itervalues()
if within:
if order:
pairs = flatten(((f1, f2), (f2, f1)) for fragments in vals
for f1, f2 in combinations(fragments, 2))
else:
pairs = (tuple(
sorted((f1, f2), key=lambda f: (f.name, f.interval.start)))
for fragments in vals
for f1, f2 in combinations(fragments, 2))
else: # across classes
if order:
pairs = (((f1, f2), (f2, f1))
for f1, f2 in combinations(flatten(vals), 2))
pairs = flatten(pairs)
else:
pairs = (tuple(
sorted((f1, f2), key=lambda f: (f.name, f.interval.start)))
for f1, f2 in combinations(flatten(vals), 2))
return unique(
ifilterfalse(lambda f: f[0].interval.overlaps_with(f[1].interval),
pairs))
def iter_pairs(self, within, order):
"""
Iterate over FragmentToken pairs.
Parameters
----------
within : bool
Only select pairs from the same class.
order : bool
Also include reverse of a pair.
Returns
-------
Iterator over (FragmentToken, FragmentToken) pairs.
"""
vals = self.clsdict.itervalues()
if within:
if order:
pairs = flatten(((f1, f2), (f2, f1))
for fragments in vals
for f1, f2 in combinations(fragments, 2))
else:
pairs = (tuple(sorted((f1, f2),
key=lambda f: (f.name, f.interval.start)))
for fragments in vals
for f1, f2 in combinations(fragments, 2))
else: # across classes
if order:
pairs = (((f1, f2), (f2, f1))
for f1, f2 in combinations(flatten(vals), 2))
pairs = flatten(pairs)
else:
pairs = (tuple(sorted((f1, f2),
key=lambda f: (f.name, f.interval.start)))
for f1, f2 in combinations(flatten(vals), 2))
return unique(ifilterfalse(lambda f: f[0].interval.overlaps_with(f[1].interval),
pairs))
def typeset(pairs):
"""
Yield the unique marks in a pair iterator.
Parameters
----------
pairs : Iterator over (FragmentToken, FragmentToken) pairs
Returns
-------
Iterator over strings
Unique marks.
"""
return unique(f.mark for f in flatten(pairs))
def typeset(pairs):
"""
Yield the unique marks in a pair iterator.
Parameters
----------
pairs : Iterator over (FragmentToken, FragmentToken) pairs
Returns
-------
Iterator over strings
Unique marks.
"""
return unique(f.mark for f in flatten(pairs))
def __init__(self, container, threshold=0.03):
self.threshold = threshold
if hasattr(container, 'iter_fragments'):
iterator = container.iter_fragments()
else:
iterator = (f for f in container)
bounds = defaultdict(list)
length = 0
for fragment in iterator:
bounds[fragment.name].append(fragment.interval)
length += 1
self.length = length
self.bounds = {}
for name, intervals in bounds.iteritems():
points = set(flatten((interval.start, interval.end)
for interval in intervals))
self.bounds[name] = np.sort(np.array(list(points)))
def unique_flatten(pairs):
r"""
Flatten a sequence of (FragmentToken, FragmentToken) pairs.
This functions yields each FragmentToken only once.
.. math::
\mathrm{flat}(P) = \{(i, j) | \exists q(((i, j), q) \in P)\}
TODO check math
Parameters
----------
pairs : Iterator over (FragmentToken, FragmentToken) pairs
Returns
-------
Iterator over FragmentTokens
"""
return unique(flatten(pairs))
def Psubs(clsdict, corpus, minlength=3, maxlength=20):
"""
Generate Psubs - the substring completion of a set of pairs.
Psubs is the association between all substrings of the pairs in classdict.
Parameters
----------
clsdict : ClassDict
corpus : Corpus
minlength : int, optional
minimum number of phones for the substrings
maxlength : int, optional
maximum number of phones for the substrings
Returns
-------
Iterator over (FragmentToken, FragmentToken) pairs
"""
sub_pairs = (pairwise_substring_completion(f1, f2, corpus, minlength,
maxlength)
for f1, f2 in clsdict.iter_pairs(within=True, order=True))
return unique(flatten(sub_pairs))
def unique_flatten(pairs):
r"""
Flatten a sequence of (FragmentToken, FragmentToken) pairs.
This functions yields each FragmentToken only once.
.. math::
\mathrm{flat}(P) = \{(i, j) | \exists q(((i, j), q) \in P)\}
TODO check math
Parameters
----------
pairs : Iterator over (FragmentToken, FragmentToken) pairs
Returns
-------
Iterator over FragmentTokens
"""
return unique(flatten(pairs))
def iter_segments(self):
return flatten(self.itervalues())
def iter_fragments(self):
return flatten(self.iter_segments())