def normalize(self, word = u""):
"""
Normalize a word.
Convert some leters forms into unified form.
@param word: the input word, if word is empty,
the word member of the class is normalized.
@type word: unicode.
@return: normalized word.
@rtype: unicode.
"""
if word == u'' and self.word == u"":
return u""
elif word != u'':
self.word = word
else:
word = self.word
self.normalized = normalize.normalize_searchtext(word)
return self.normalized
from pyquery import PyQuery as pq
import tashaphyne.normalize as norm
import Levenshtein
almizan = codecs.open('data/output-convert.html', encoding='utf-8').read()
quran = codecs.open('data/quran.txt', encoding='utf-8').readlines()
errors = open('data/errors.txt','w')
d = pq(almizan)
for i, aya in enumerate(d('blockquote p')):
aya = pq(aya)
"""if aya.parent().prev()[0].tag != 'h1':
continue;"""
s, a, max, c = 0, 0, 0, 0
for ayah in quran:
ayah = norm.normalize_searchtext(ayah)
aya1 = norm.normalize_searchtext(aya[0].text)
if max < Levenshtein.ratio(ayah,aya1):
max = Levenshtein.ratio(ayah,aya1)
if Levenshtein.ratio(ayah, aya1) > 0.75:
match = re.search(r'\d+', ayah)
s = int(match.group(0))
match = re.search(r'\|\d+', ayah)
a = int(match.group(0)[1:])
print('L: %s-%s' %(s,a))
for err in quran[:c]:
errors.write(err.encode('utf-8'))
del quran[:c+1]
break
c += 1
if max <= 0.75:
while(i <= aya_end):
aya_num = str(sura) +'|'+ str(i)
for c in range(len(quran)):
q_aya = quran[c]
if(q_aya.startswith(aya_num)):
sec_ayas.append(q_aya)
quran = quran[c+1:]
break
i += 1
counter = 1
for part in sec.find("em"):
success = 0
part = pq(part)
try:
part_text = norm.normalize_searchtext(part[0].text)
part_text = isri.stem(part_text)
except:
part_text = part[0].text
try:
errors.write(part.outerHtml())
except:
pass
print counter
counter += 1
if part == None or part[0].text == None:
continue
if len(part[0].text)<=2:
continue
for aya in sec_ayas:
from nltk.stem import arlstem
from nltk.corpus import stopwords
from tashaphyne import normalize
data = pd.read_excel('../feature datasets/ar/labels.xlsx', encoding = 'utf-8', index_col = 0)
# Tokenize / Filter / Normalize
for index, row in data.iterrows():
# Tokenization
comment_tokens = re.split("[;., \-!?:\*]+", row['text'])
#stemmer = arlstem.ARLSTem()
#comment_tokens = list(map(stemmer.stem, comment_tokens))
# Filtering
filtered_words = [word for word in comment_tokens if word not in stopwords.words('arabic')]
filtered_words = [re.sub('[^\u0621-\u0652]+', '', i) for i in filtered_words]
filtered_words = list(filter(None, filtered_words))
# Stemming
#stemmer = arlstem.ARLSTem()
#stemmed_words = list(map(stemmer.stem, filtered_words))
# Normalization
normalized = normalize.normalize_searchtext(' '.join(filtered_words))
data.loc[index, 'text'] = normalized
data['text'].replace([''], np.nan, inplace=True)
clean_data = data.dropna()
clean_data.to_excel('../feature datasets/ar/cleaned_data.xlsx', index_label="index")
