def generate(division):
if division=='ns':
each(cat, cross([northern, southern], 'path trigram'.split()))
else:
ns = northern[1] + southern[1]
random.shuffle(ns)
n = ('NorthernRandom', ns[:len(northern[1])])
s = ('SouthernRandom', ns[len(northern[1]):])
each(cat, cross([n, s], 'path trigram'.split()))
def groupavg((c1,c2)):
"group average"
return avg(map(compose(edges.__getitem__, frozenset),
cross(flatten(c1), flatten(c2))))
def complete((c1,c2)):
"complete link"
return max(map(compose(edges.__getitem__, frozenset),
cross(flatten(c1), flatten(c2))))
def single((c1,c2)):
"single link"
return min(map(compose(edges.__getitem__, frozenset),
cross(flatten(c1), flatten(c2))))
def sed_avg(ws1, ws2):
"[{str:[float]}]*[{str:[float]}] -> float"
segs1,segs2 = (concat(transpose_word(ws1)), concat(transpose_word(ws1)))
return lst.avg(map(fnc.uncurry(feature_sub), lst.cross(segs1, segs2)))
avgregions = lst.avg(map(sed_avg_total, regions))
return dict(zip(keys, map(sed_distance(avgregions), regions)))
def feature_sub(seg1, seg2):
"({str:float}*{str:float}) -> float"
return (len(set(seg1) ^ set(seg2))
+ sum(abs(f1-f2) for f1,f2 in dct.zip(seg1,seg2).values()))
@curried
def sed_distance(avg, (region1, region2)):
"float*([[{str:[float]}]],[[{str:[float]}]])->float"
return sum(map(sed_levenshtein(avg), zip(region1, region2)))
def transpose_word(word):
"[{str:[float]}] -> [[{str:float}]]"
def transpose_segment(seg):
return [dict(zip(seg.keys(), ns)) for ns in lst.transpose(seg.values())]
return lst.transpose(map(transpose_segment, word))
@curried
def sed_levenshtein(avg,(ws1,ws2)):
"float*([{str:[float]}],[{str:[float]}])->float"
def levenshtein((w1, w2)):
return lev._levenshtein(w1, w2, avg,
(lambda _:avg,lambda _:avg,feature_sub))[-1][-1]
return lst.avg(map(levenshtein,
lst.cross(transpose_word(ws1), transpose_word(ws2))))
def sed_avg(ws1, ws2):
"[{str:[float]}]*[{str:[float]}] -> float"
segs1,segs2 = (concat(transpose_word(ws1)), concat(transpose_word(ws1)))
return lst.avg(map(fnc.uncurry(feature_sub), lst.cross(segs1, segs2)))
def sed_avg_total((region1, region2)):
"([[{str:[float]}]],[[{str:[float]}]]) -> float"
return lst.avg(map(sed_avg, region1, region2)) / 2
