def compare(c1,c2,similarity):
from needleman_wunsch import align
alignment = align(c1,c2,S=similarity)
if alignment[0][0] == 'NA': return 'NA'
c = []
for i in range(0,len(alignment[0])):
if alignment[0][i][0] in ['[',']'] and alignment[0][i][0] != alignment[1][i][0]:
return 'NA'
if alignment[0][i][0] == alignment[1][i][0] in ['[',']'] and (alignment[0][i][1] != alignment[1][i][1]):
c.append([alignment[0][i][0], 'XP'])
elif alignment[0][i][1] == alignment[1][i][1]:
if alignment[0][i][0] == alignment[1][i][0]:
c.append(alignment[0][i])
else: c.append(['*', alignment[0][i][1]])
else: c.append(['*','*'])
context = [c[0]]
for i in range(1,len(c)):
if c[i][1] != '*' or c[i-1][1] != '*': context.append(c[i])
open_nodes, closed_nodes = [], []
hasGap = False
for word in context:
if word[0] == '[': open_nodes.append(word[1])
if word[0] == ']':
if len(open_nodes) == 0: return 'NA'
elif open_nodes[len(open_nodes)-1] == word[1]:
open_nodes = open_nodes[0:len(open_nodes)-1]
else: return 'NA'
if word[1] == 'VB____': hasGap = True
if not hasGap or len(open_nodes)>1: return 'NA'
return context
def match_kana(kanji, kana, return_score=False):
def stoponzero(alignments):
for i, alignment in enumerate(alignments):
furigana, score = finalize_furigana(alignment, return_score=True)
yield furigana, score
if score == 0:
break
best_match, score = min(stoponzero(filter_alignments(align(kanji, kana))),
key=lambda x: x[1])
return (best_match, score) if return_score else best_match
def approximate_split(readings, kana, max_distance):
def similarity(a,b):
if a==b:
return 1
if a==' ' or b==' ':
return -10
return -1
for result, score in readings:
if score > max_distance:
break
q = kana
w = ' '.join([y for _,y in result])
l, r = next(align(q, w, d=-1, fill=" ", s=similarity))
res = filter(None, l.split(' '))
res = list(zip([x for x,_ in result], res))
if len(res) == len(result):
return res