# ---------------- select_features_by_positive_degree() ----------------
def select_features_by_positive_degree(tsm_positive, tsm_unlabeled, (threshold_pd_word, threshold_speciality, threshold_popularity)):
total_terms = tsm_positive.get_total_terms()
total_samples = tsm_positive.get_total_samples()
selected_terms = {}
rowidx = 0
logging.debug(Logger.debug("Calculate PDword. %d samples, %d terms in tsm_positive" % (total_samples, total_terms)))
for term_id in tsm_positive.term_matrix():
term_info = tsm_positive.get_term_row(term_id)
(pd_word, speciality, popularity) = calculate_term_positive_degree(term_id, tsm_positive, tsm_unlabeled, None)
if pd_word >= threshold_pd_word and speciality >= threshold_speciality and popularity >= threshold_popularity:
selected_terms[term_id] = (pd_word, speciality, popularity)
if rowidx % 1000 == 0:
logging.debug(Logger.debug("feature_selection() %d/%d - pd_word:%.6f speciality:%.6f popularity:%.6f" % (rowidx, total_terms, speciality + popularity, speciality, popularity)))
rowidx += 1
return selected_terms
# ---------------- get_terms_positive_degree_by_category() ----------------
def get_terms_positive_degree_by_category(tsm, positive_samples_list, unlabeled_samples_list):
tsm_positive = tsm.clone(positive_samples_list)
tsm_unlabeled = tsm.clone(unlabeled_samples_list)
def query_by_id(self, samples_positive, samples_unlabeled, sample_id):
tsm_positive = samples_positive.tsm
tsm_unlabeled = samples_unlabeled.tsm
sensitive_words = {
##u"立案":3.0,
##u"获刑":3.0,
##u"受贿":3.0,
##u"有期徒刑":3.0,
##u"宣判":3.0,
##u"审计":2.0,
##u"调查":2.0
}
sensitive_terms = self.transform_sensitive_terms(sensitive_words, self.vocabulary)
try:
sample_content = samples_unlabeled.db_content.Get(str(sample_id))
#(_, category, date, title, key, url, content) = msgpack.loads(sample_content)
(_, category, date, title, key, url, msgext) = decode_sample_meta(sample_content)
(version, content, (cat1, cat2, cat3)) = msgext
print "sample id: %d" % (sample_id)
print "category: %d" % (category)
print "key: %s" % (key)
print "url: %s" % (url)
print "date: %s" % (date)
print "title: %s" % (title)
print "---------------- content ----------------"
#print "%s" % (content)
sample_terms, term_map = self.vocabulary.seg_content(content)
print "sample_terms: %d terms_count: %d" % (sample_terms, len(term_map))
#for term_id in term_map:
terms_list = sorted_dict_by_values(term_map, reverse=True)
for (term_id, term_used_in_sample) in terms_list:
term_text = self.vocabulary.get_term_text(term_id)
#term_used_in_sample = term_map[term_id]
print "%s(%d): %d" % (term_text, term_id, term_used_in_sample)
except KeyError:
print "Sample %d not found in db_content." % (sample_id)
db_sm = samples_unlabeled.tsm.open_db_sm()
try:
str_sample_info = db_sm.Get(str(sample_id))
(category, sample_terms, term_map) = msgpack.loads(str_sample_info)
print ""
print "---------------- keywords ----------------"
print ""
terms = {}
for term_id in term_map:
term_text = self.vocabulary.get_term_text(term_id)
term_used = term_map[term_id]
(pd_word, speciality, popularity) = calculate_term_positive_degree(term_id, tsm_positive, tsm_unlabeled, sensitive_terms)
terms[term_id] = (pd_word, speciality, popularity, term_used, term_text)
terms_list = sorted_dict_by_values(terms, reverse = True)
for (term_id, (pd_word, speciality, popularity, term_used, term_text)) in terms_list:
print "%s\t%d\t[%.6f,%.6f,%.6f]\t(id:%d)" % (term_text, term_used, pd_word, speciality, popularity, term_id)
except KeyError:
print "Sample %d not found in db_sm." % (sample_id)
samples_unlabeled.tsm.close_db(db_sm)