def langsim(fname, lang, threshold, phon, only_hr=False, topk=20):
"""
Gets a topk list of languages similar to this language, various parameters control this.
:param fname:
:param lang:
:param threshold:
:param phon:
:param only_hr:
:param topk:
:return:
"""
langs = wals.loadLangs(fname)
#langs = filter(lambda l: l.nonzerofrac > threshold, langs)
tgtlang = None
for l in langs:
if l["iso_code"].decode("utf8") == lang:
tgtlang = l
break
if tgtlang == None:
return [(-1, "Language '{0}' not found...".format(lang.encode("utf8")))]
# now filter by high resource
# get tgtlang first, b/c it may not be high resource
if only_hr:
langs = filter(lambda l: l.hr, langs)
dists = []
tgtf = tgtlang.phon_feats()
for l in langs:
if l["iso_code"].decode("utf8") == lang:
continue
t = l.phon_feats()
#dist = cosine(tgtf, t)
numequal = sum(np.equal(tgtf, t))
# now remove all places where they are both zero
a = set(np.where(tgtf == 0)[0])
b = set(np.where(t == 0)[0])
numequal -= len(a.intersection(b))
dist = 1 - numequal / float(len(tgtf))
dists.append((dist, l.fullname()))
dists = sorted(dists)
return dists[:topk]
def compareFeats(fname, lang1, lang2):
langs = wals.loadLangs(fname)
l1 = None
l2 = None
for lang in langs:
if lang["Name"].decode("utf8") == lang1:
l1 = lang
elif lang["Name"].decode("utf8") == lang2:
l2 = lang
return l1,l2