def cal_q_d_log(self, q, dc=None, d=None, clean=False, content_field='content'):
'''
:param q: 这个q可以是string句子，可以是wordlist，这个我来处理，
:param dc:
:param d: d是{}类型，d[content_field]表示的是内容，这个内容是做过了基本处理的内容
:param clean:
:param content_field:
:return:内容是P(w|d)
'''
if type(q) != list:
q = basic_preprocess(q)
if type(d[content_field]) != list:
d[content_field] = basic_preprocess(d[content_field][0])
if clean: q = clean_tools(q)
if dc is None and d is not None:
if clean:
d[content_field] = clean_tools(d[content_field])
dc = defaultdict(int)
for w in d[content_field]:
dc[w] += 1.0
# print '[DEBUG] dc:', dc['US'], dc
ret = 1.0
for w in q:
ret += math.log((dc[w] + self.lmd * self.C[w]) / float( len(d[content_field]) + self.lmd )) #self.lmd * self.C[w] + (1-self.lmd) * dc[w]
return ret
def test_2():
q = 'US Military Crisis Response'
q = cut_words(q)
d1 = sample_doc # 'US Military Crisis Response US Military Crisis Response US Military Crisis Response'
d1 = basic_preprocess(d1)
d2 = {
'score': 9.644032,
'key': 'ebola-1bbd62fe484a96be675ab80a304f0320a742a2da67f696cde413aee99e9f9349',
'content': d1
}
lm = LMDirichlet()
dc = lm.cal_dc(d1)
lm.load(LMDirichlet_Json)
print 'lm.C[US]:', lm.C['US']
p1 = lm.cal_w_d('US', d1)
p2 = lm.cal_q_d(q, d=d2)
p3 = lm.cal_q_d_log(q, d=d2)
p4 = lm.cal_q_d(q, d=d2, dc=dc)
print "p w,d:", p1, math.log(p1)
print "p q,d:", p2
print "p q,d:", p3
print "p d d by dc:", p4
def test_3():
from data_utils import basic_preprocess
q = 'US Military Crisis Response'
q = cut_words(q)
d1 = sample_doc # 'US Military Crisis Response US Military Crisis Response US Military Crisis Response'
q_list = basic_preprocess(q)
d1 = basic_preprocess(d1)
d2 = {
'score': 9.644032,
'key': 'ebola-1bbd62fe484a96be675ab80a304f0320a742a2da67f696cde413aee99e9f9349',
'content': d1
}
lm = LMDirichlet()
dc = lm.cal_dc(d1)
#LMDirichlet_without_stem
lm.load(LMDirichlet_Json)
# print 'lm.C[US]:', lm.C['US']
# print d1
p1 = lm.cal_w_d('US', d1)
p2 = lm.cal_q_d(q, d=d2)
p3 = lm.cal_q_d_log(q, d=d2)
p4 = lm.cal_q_d(q, d=d2, dc=dc)
print "p w,d:", p1, math.log(p1)
print "p q,d:", p2
print "p q,d:", p3
print "p d d by dc:", p4
print "===================="
print "q list:", q_list
lm.load(LMDirichlet_without_stem)
# print 'lm.C[US]:', lm.C['US']
p1 = lm.cal_w_d(q_list[0], d1)
p2 = lm.cal_q_d(q, d=d2)
p3 = lm.cal_q_d_log(q, d=d2)
p4 = lm.cal_q_d(q, d=d2, dc=dc)
print "p w,d:", p1, math.log(p1)
print "p q,d:", p2
print "p q,d:", p3
print "p d d by dc:", p4
def cal_q_d(self, q, dc=None, d=None, clean=False, content_field='content'):
'''
:param q: 必须是word list
:param dc:
:param d: d是{}类型，d[content_field]表示的是内容，这个内容是做过了基本处理的内容
:param clean:
:param content_field:
:return:内容是P(w|d)
'''
# print "QUERY:", q
if type(q) != list:
print "before process q:", q
q = basic_preprocess(q)
print "after process q:", q
if type(d[content_field]) != list:
d[content_field] = basic_preprocess(d[content_field][0])
if clean: q = clean_tools(q)
if dc is None and d is not None:
if clean:
d[content_field] = clean_tools(d[content_field])
dc = defaultdict(int)
for w in d[content_field]:
dc[w] += 1.0
# print 'CHECKING..., doc len:', len(d[content_field])
# for w in q:
# print "check w cnt:", w, dc[w]
# print '[DEBUG] dc:', dc['US'], dc
ret = 1.0
for w in q:
ret *= (dc[w] + self.lmd * self.C[w]) / (
float(len(d[content_field]) + self.lmd) # * self.C[w] # * len(d[content_field])
)
# ret *= (dc[w] + self.lmd * self.C[w]) / (
# float(len(d[content_field]) + self.lmd)
# ) #self.lmd * self.C[w] + (1-self.lmd) * dc[w]
return ret
def deal_nytimes(file_id,
in_dir,
out_dir,
json_dir,
stem_dir,
stem_jsdir,
overwrite=False):
def get_content(blocks):
try:
max_len = max([len(b) for b in blocks])
main = [b for b in blocks if len(b) == max_len][0].strip()
blocks = [b for b in blocks if main.startswith(b.strip())]
return blocks + [main]
except Exception as e:
logging.exception("[!][%s] Exception: %s", file_id, e)
return []
def write_line_data(file_name, key, words):
with codecs.open(file_name, "w", "utf-8") as fl:
fl.write("{} ".format(key))
fl.write(",".join(words))
fl.write("\n")
in_file = in_dir.format(file_id)
out_file = out_dir.format(file_id)
json_file = json_dir.format(file_id)
stem_file = stem_dir.format(file_id)
stem_json = stem_jsdir.format(file_id)
if exists(out_file) and exists(stem_file) and not overwrite:
return False
logging.info("[#] dealing file: %s", in_file)
js = json.load(codecs.open(in_file, "r", "utf-8"))
text = " ".join([js["title"]] + get_content(js["content"].values()))
words = du.basic_preprocess(text)
if not exists(out_file) or overwrite:
write_line_data(out_file, js["doc_id"], words)
stems = du.stemmer_by_porter(words)
if not exists(stem_file) or overwrite:
write_line_data(stem_file, js["doc_id"], stems)
doc_dict = {"words": Counter(words), "id": file_id, "key": js["doc_id"]}
if not exists(json_file) or overwrite:
json.dump(doc_dict, codecs.open(json_file, "w", "utf-8"))
if not exists(stem_json) or overwrite:
doc_dict["words"] = Counter(stems)
json.dump(doc_dict, codecs.open(stem_json, "w", "utf-8"))
return True
def deal_ebola(file_id,
in_dir,
out_dir,
json_dir,
stem_dir,
stem_jsdir,
overwrite=True):
def write_line_data(file_name, key, words):
with codecs.open(file_name, "w", "utf-8") as fl:
fl.write("{} ".format(key))
fl.write(",".join(words))
fl.write("\n")
in_file = in_dir.format(file_id)
out_file = out_dir.format(file_id)
json_file = json_dir.format(file_id)
stem_file = stem_dir.format(file_id)
stem_json = stem_jsdir.format(file_id)
logging.info("[#] deal file_id: %s", file_id)
if (exists(out_file) and not overwrite and exists(stem_file)
and exists(json_file) and exists(stem_json)):
return False
logging.info("[#] dealing file_id: %s", file_id)
js = json.load(codecs.open(in_file, "r", "utf-8"))
text = parse_html(js["content"])
words = du.basic_preprocess(text, length_limit=1)
doc_dict = {"words": Counter(words), "id": file_id, "key": js["key"]}
stem_words = du.stemmer_by_porter(words)
if not exists(json_file) or overwrite:
json.dump(doc_dict, codecs.open(json_file, "w", "utf-8"))
if not exists(out_file) or overwrite:
write_line_data(out_file, js["key"], words)
if not exists(stem_file) or overwrite:
write_line_data(stem_file, js["key"], stem_words)
if not exists(stem_json) or overwrite:
doc_dict["words"] = Counter(stem_words)
json.dump(doc_dict, codecs.open(stem_json, "w", "utf-8"))
return True
