def proc_right_segments(relation, segments):
seg_list = proc_line(segments, ';')
word_list = []
pos_list =[]
for seg in seg_list:
if seg == '': continue
sp_list = proc_line(seg, ':')
word = sp_list[0]
pos = sp_list[1]
word_list.append(word)
pos_list.append(pos)
total_featureset_dict[relation]['right_words'].add(word)
total_featureset_dict[relation]['p2_right_pos'].add(pos)
featureset_dict[relation]['right_words'].append(word_list)
featureset_dict[relation]['p2_right_pos'].append(pos_list)
def gen_total_featureset(relation):
filename = trainset_prefix + relation
file_in = os.path.join(data_dir, filename)
fi = open(file_in.decode('utf-8'), 'r')
for line in fi:
seg_list = proc_line(line, '||')
proc_left_segments(relation, seg_list[0])
proc_middle_segments(relation, seg_list[1])
proc_right_segments(relation, seg_list[2])
fi.close()
def gen_training_corpus(text):
seg_list = proc_line(text, "\t")
relation = seg_list[0].strip()
person1 = seg_list[1].strip()
person2 = seg_list[2].strip()
title = seg_list[3].strip()
has_rel = seg_list[4].strip() # 是否的确有对应关系
if has_rel == "0":
relation = "null"
label = fetch_label(relation)
return corpus(title, person1, person2, label)
def read_extracted_featureset():
for feature_name in feature_name_list:
filename = extracted_featset_prefix + feature_name
file_in = os.path.join(featureset_dir, filename)
fi = open(file_in, 'r')
line = fi.readline()
seg_list = proc_line(line, ';')
featset = []
for seg in seg_list:
if seg == '': continue
featset.append(seg)
extracted_featureset_dict[feature_name] = featset
fi.close()
