def phono_edit_distance(word1, word2, sequence_type, features):
"""Returns an analogue to Levenshtein edit distance but uses
phonological features instead of characters
Parameters
----------
word1: Word
Word object containing transcription tiers which will be compared
to another word containing transcription tiers
word2: Word
The other word containing transcription tiers to which word1 will
be compared
sequence_type: string
Name of the sequence type (transcription or a tier) to use for comparisons
features: FeatureMatrix
FeatureMatrix that contains all the segments in both transcriptions
to be compared
Returns
-------
float
the phonological edit distance between two words
"""
w1 = getattr(word1, sequence_type)
w2 = getattr(word2, sequence_type)
a = Aligner(features_tf=True, features=features)
m = a.make_similarity_matrix(w1, w2)
return m[-1][-1]['f']
def find_mutation_minpairs(corpus_context,
query,
tier_type=None,
stop_check=None,
call_back=None):
"""Find all minimal pairs of the query word based only on segment
mutations (not deletions/insertions)
Parameters
----------
corpus_context : CorpusContext
Context manager for a corpus
query : Word
The word whose minimal pairs to find
stop_check : callable or None
Optional function to check whether to gracefully terminate early
call_back : callable or None
Optional function to supply progress information during the function
Returns
-------
list
The found minimal pairs for the queried word
"""
matches = []
sequence_type = corpus_context.sequence_type
query = ensure_query_is_word(query, corpus_context,
corpus_context.sequence_type, tier_type)
if call_back is not None:
call_back('Finding neighbors...')
call_back(0, len(corpus_context))
cur = 0
al = Aligner(features_tf=False,
ins_penalty=float('inf'),
del_penalty=float('inf'),
sub_penalty=1)
for w in corpus_context:
w_sequence = getattr(w, sequence_type)
query_sequence = getattr(query, sequence_type)
if stop_check is not None and stop_check():
return
if call_back is not None:
cur += 1
if cur % 10 == 0:
call_back(cur)
if (len(w_sequence) > len(query_sequence) + 1
or len(w_sequence) < len(query_sequence) - 1):
continue
m = al.make_similarity_matrix(query_sequence, w_sequence)
if m[-1][-1]['f'] != 1:
continue
matches.append(str(w_sequence))
neighbors = list(set(matches) - set([str(query_sequence)]))
return (len(neighbors), neighbors)
def find_mutation_minpairs(corpus_context, query, tier_type = None, collapse_homophones = False,
stop_check = None, call_back = None):
"""Find all minimal pairs of the query word based only on segment
mutations (not deletions/insertions)
Parameters
----------
corpus_context : CorpusContext
Context manager for a corpus
query : Word
The word whose minimal pairs to find
stop_check : callable or None
Optional function to check whether to gracefully terminate early
call_back : callable or None
Optional function to supply progress information during the function
Returns
-------
list
The found minimal pairs for the queried word
"""
matches = []
sequence_type = corpus_context.sequence_type
query = ensure_query_is_word(query, corpus_context, corpus_context.sequence_type, tier_type)
if call_back is not None:
call_back('Finding neighbors...')
call_back(0,len(corpus_context))
cur = 0
al = Aligner(features_tf=False, ins_penalty=float('inf'), del_penalty=float('inf'), sub_penalty=1)
for w in corpus_context:
w_sequence = getattr(w, sequence_type)
query_sequence = getattr(query, sequence_type)
if stop_check is not None and stop_check():
return
if call_back is not None:
cur += 1
if cur % 10 == 0:
call_back(cur)
if (len(w_sequence) > len(query_sequence)+1 or
len(w_sequence) < len(query_sequence)-1):
continue
m = al.make_similarity_matrix(query_sequence, w_sequence)
if m[-1][-1]['f'] != 1:
continue
w_sequence = getattr(w, sequence_type)
if collapse_homophones and any(getattr(m, sequence_type) == w_sequence for m in matches):
continue
else:
#matches.append(str(w_sequence))
matches.append(w)
matches = [m.spelling for m in matches]
neighbors = list(set(matches)-set([str(query_sequence)]))
return (len(neighbors), neighbors)
def phono_edit_distance(word1, word2, sequence_type, features):
"""Returns an analogue to Levenshtein edit distance but uses
phonological features instead of characters
Parameters
----------
word1: Word
Word object containing transcription tiers which will be compared
to another word containing transcription tiers
word2: Word
The other word containing transcription tiers to which word1 will
be compared
sequence_type: string
Name of the sequence type (transcription or a tier) to use for comparisons
features: FeatureMatrix
FeatureMatrix that contains all the segments in both transcriptions
to be compared
Returns
-------
float
the phonological edit distance between two words
"""
w1 = getattr(word1,sequence_type)
w2 = getattr(word2,sequence_type)
a = Aligner(features_tf=True, features=features)
m = a.make_similarity_matrix(w1, w2)
return m[-1][-1]['f']
