def wd_seg(self):
def _file_seg_(filename):
_f = codecs.open(filename, 'r', 'utf-8')
txt = _f.read()
seg_list = [_wd for _wd in jieba.cut(txt)]
# words account before delete stopwords
_stopwd = len(seg_list)
for sw in self.stop_wd_list:
while sw.strip() in seg_list:
seg_list.remove(sw.strip())
# words account before after stopwords
stopwd_ = len(seg_list)
_f.close()
return seg_list, _stopwd, stopwd_
# use get_sets_of_root_path_tree so
for iter_filename in self.set_dict['files']:
log_dir = '\\'.join(iter_filename.split('.')[0].split('\\')[:-3]) + '-seg\\' + \
'\\'.join(iter_filename.split('.')[0].split('\\')[-3:-1])
if not os.path.exists(log_dir):
os.makedirs(log_dir)
log_name = iter_filename.split('.')[0].split('\\')[-1] + '.log'
wd_list, _wds, wds_ = _file_seg_(iter_filename)
Log.log_blue_running("Processed " + self.set_dict['category'] + ' - ' + iter_filename + ',' +
' words account reduced from ' + str(_wds) + ' to ' + str(wds_))
f = codecs.open(log_dir + '\\' + log_name, 'w', 'utf-8')
for wd in wd_list:
if wd != u' ' and wd != u'\n':
f.write(wd)
f.write(' ')
f.close()
def text_classification_polynomial(self, test_category):
logfile = open('tc_log.log', 'a')
prior_pos = self._get_prior_possibility_()
def p_possibility(p_pos_list):
log_p_pos = map(math.log, p_pos_list)
log_p = reduce(lambda x, y: x + y, log_p_pos)
return log_p
def classification(p_possibility_list, category):
p = max([
p_possibility_list[_iter_]
for _iter_ in range(len(p_possibility_list))
]).values()[0]
return p, p == category
test_num = len(self._test_set_dict_[test_category])
if_right = 0
for _iter_file in self._test_set_dict_[test_category]:
wd_list = self.get_wd_list(_iter_file)
category_pos = []
for _iter_category in range(len(self._wd_list_category_)):
wd_account = reduce(
lambda x, y: x + y,
self._wd_list_category_[_iter_category]['words'].values())
_p_category = []
denominator = wd_account + len(self._wd_list_total_)
for wd in wd_list.keys():
if wd in self._wd_list_category_[_iter_category]['words']:
nij = self._wd_list_category_[_iter_category]['words'][
wd]
else:
nij = 0
# every single word count for once
for _iter_account_ in range(wd_list[wd]):
_p = (nij + 1) * 1.0 / denominator
_p_category.append(_p)
category_pos.append({
p_possibility(_p_category) + math.log(prior_pos[_iter_category]['prior_possibility']):
self._wd_list_category_[_iter_category]['category']
})
if classification(category_pos, test_category)[1]:
if_right += 1
Log.log_blue_running('Category: ' + test_category + ' total ' +
str(test_num) + ', correct: ' + str(if_right) +
' Accuracy: ' +
str((if_right * 1.0 / test_num) * 100) + '%')
logfile.write('Category: ' + test_category + ' total ' +
str(test_num) + ', correct: ' + str(if_right) +
' Accuracy: ' + str((if_right * 1.0 / test_num) * 100) +
'%' + '\n')
logfile.close()
def _bayes_train_(self):
self._wd_list_category_ = []
if os.path.exists(self._cache_path_total_):
Log.log_blue_running('Corpus cache loaded.')
self._wd_list_total_ = self.load_cache(self._cache_path_total_)
for _iter_cache_file in range(len(self._cache_path_category_)):
self._wd_list_category_.append({
'category':
self._cache_path_category_[_iter_cache_file]['category'],
'words':
self.load_cache(
self._cache_path_category_[_iter_cache_file]
['cache_path'])
})
else:
Log.log_blue_running('Naive Bayes train module.')
wd_list_total = {}
for _iter_basename in self._train_set_dict_:
characteristic_wd_category = {}
for _iter_filename in self._train_set_dict_[_iter_basename]:
tmp = TextClassification.get_wd_list(_iter_filename)
for wd in tmp:
if wd in wd_list_total:
wd_list_total[wd] += tmp[wd]
else:
wd_list_total[wd] = tmp[wd]
words_freq = OrderedDict(
(sorted((copy.deepcopy(tmp).items()),
key=lambda t: -t[-1]))).items()
if len(words_freq) > 10:
characteristic_wd_file = dict(
(x, y) for x, y in words_freq[:10])
else:
characteristic_wd_file = dict(
(x, y) for x, y in words_freq)
for wd in characteristic_wd_file:
if wd in characteristic_wd_category:
characteristic_wd_category[
wd] += characteristic_wd_file[wd]
else:
characteristic_wd_category[
wd] = characteristic_wd_file[wd]
self.create_cache(
characteristic_wd_category,
self._train_set_dir_ + '\\' + _iter_basename + '.json')
Log.log_blue_running('Category: ' + _iter_basename +
' cache created.')
self._wd_list_total_ = wd_list_total
self.create_cache(
self._wd_list_total_,
os.path.abspath(self._train_set_dir_ + '\\..') +
'\\wdlist-total.json')
Log.log_blue_running('Corpus word-lists cache created.')
def get_stop_wd_list():
with codecs.open('stopwd.txt', 'r', 'utf-8') as f:
text = f.readlines()
stop_wd_list = [word for word in text]
Log.log_blue_running('Stopwords total : ' + str(len(stop_wd_list)) + '\n')
return stop_wd_list
