def get_inquirer_vec(relation, Arg, parse_dict):
inquirer = Non_Explicit_dict().inquirer
inquirer_stem = Non_Explicit_dict().inquirer_stem
list = [0] * 42
verb_tag= ["MD", "VB", "VBD", "VBG", "VBN", "VBP", "VBZ"]
# ["NN/dog" ,"NNS/joks]
word_list = _get_Arg_word_pos_list(relation, Arg, parse_dict)
for item in word_list:
if item == "":
continue
tag, word = item.split("/")[:2]
if tag not in verb_tag:
continue
word = word.lower()
if word in inquirer.keys():
list = _merge(list, inquirer[word])
else:
stem = util.stem_string(word)
if stem in inquirer.keys():
list = _merge(list, inquirer[stem])
elif stem in inquirer_stem.keys():
list = _merge(list, inquirer_stem[stem])
return list
def is_negate(wordlist):
negate = Non_Explicit_dict().negate
negate_stem = Non_Explicit_dict().negate_stem
for word in wordlist:
if word in negate:
return True
else:
stem = util.stem_string(word)
if stem in negate or stem in negate_stem:
return True
return False
def firstlast_first3(relation, parse_dict):
# load dict
dict_Arg1_first = Non_Explicit_dict().dict_Arg1_first
dict_Arg1_last = Non_Explicit_dict().dict_Arg1_last
dict_Arg2_first = Non_Explicit_dict().dict_Arg2_first
dict_Arg2_last = Non_Explicit_dict().dict_Arg2_last
dict_Arg1_first_Arg2_first = Non_Explicit_dict().dict_Arg1_first_Arg2_first
dict_Arg1_last_Arg2_last = Non_Explicit_dict().dict_Arg1_last_Arg2_last
dict_Arg1_first3 = Non_Explicit_dict().dict_Arg1_first3
dict_Arg2_first3 = Non_Explicit_dict().dict_Arg2_first3
''' feature '''
Arg1_first, Arg1_last, Arg2_first, Arg2_last,\
Arg1_first_Arg2_first, Arg1_last_Arg2_last,\
Arg1_first3, Arg2_first3 \
= dict_util.get_firstlast_first3(relation, parse_dict)
features = []
features.append(get_feature_by_feat(dict_Arg1_first,Arg1_first))
features.append(get_feature_by_feat(dict_Arg1_last,Arg1_last))
features.append(get_feature_by_feat(dict_Arg2_first,Arg2_first))
features.append(get_feature_by_feat(dict_Arg2_last,Arg2_last))
features.append(get_feature_by_feat(dict_Arg1_first_Arg2_first,Arg1_first_Arg2_first))
features.append(get_feature_by_feat(dict_Arg1_last_Arg2_last,Arg1_last_Arg2_last))
features.append(get_feature_by_feat(dict_Arg1_first3,Arg1_first3))
features.append(get_feature_by_feat(dict_Arg2_first3,Arg2_first3))
return util.mergeFeatures(features)
def word2vec_cluster_pair(relation, parse_dict):
''' load dict '''
dict_word2vec_cluster_pairs = Non_Explicit_dict().dict_word2vec_cluster_pairs
''' feature '''
word2vec_cluster_pairs = dict_util.get_word2vec_cluster_pairs(relation, parse_dict)
return get_feature_by_feat_list(dict_word2vec_cluster_pairs, word2vec_cluster_pairs)
def verb_pair(relation, parse_dict):
# load dict
dict_verb_pair = Non_Explicit_dict().dict_verb_pair
# feature
verb_pair = dict_util.get_verb_pair(relation, parse_dict)
return get_feature_by_feat_list(dict_verb_pair, verb_pair)
def arg2_tense(relation, parse_dict):
# load dict
dict_arg2_tense = Non_Explicit_dict().dict_arg2_tense
''' feature '''
arg2_tense = dict_util.get_arg2_tense(relation, parse_dict)
return get_feature_by_feat(dict_arg2_tense, arg2_tense)
def arg2_first3_conn(relation, parse_dict):
# load dict
dict_arg2_first3_conn = Non_Explicit_dict().dict_arg2_first3_conn
''' feature '''
arg2_first3_conn = dict_util.get_arg2_first3_conn(relation, parse_dict)
return get_feature_by_feat(dict_arg2_first3_conn, arg2_first3_conn)
def prev_next_context_conn(relation, parse_dict, implicit_context_dict):
# load dict
dict_prev_next_context_conn = Non_Explicit_dict().dict_prev_next_context_conn
# feature
prev_next_context_conn = dict_util.get_prev_next_context_conn(relation, parse_dict, implicit_context_dict)
return get_feature_by_feat(dict_prev_next_context_conn, prev_next_context_conn)
def brown_cluster_pair(relation, parse_dict):
''' load dict '''
dict_word_pairs = Non_Explicit_dict().dict_brown_cluster
''' feature '''
brown_cluster_pairs = dict_util.get_brown_cluster_pairs(relation, parse_dict)
return get_feature_by_feat_list(dict_word_pairs, brown_cluster_pairs)
def main_verb_pair(relation, parse_dict):
# load dict
dict_main_verb_pair = Non_Explicit_dict().dict_main_verb_pair
#feature
main_verb_pair = dict_util.get_main_verb_pair(relation, parse_dict)
return get_feature_by_feat(dict_main_verb_pair, main_verb_pair)
def verbs(relation, parse_dict):
#load dict
dict_verb_classes = Non_Explicit_dict().dict_verb_classes
'''feature'''
# 1. the number of pairs of verbs in Arg1 and Arg2 from same verb class
Arg1_words = dict_util.get_Arg_Words_List(relation, "Arg1", parse_dict)
Arg2_words = dict_util.get_Arg_Words_List(relation, "Arg2", parse_dict)
count = 0
for w1, w2 in [(w1.lower(), w2.lower()) for w1 in Arg1_words for w2 in Arg2_words]:
if w1 in dict_verb_classes and w2 in dict_verb_classes:
c1 = dict_verb_classes[w1]
c2 = dict_verb_classes[w2]
if set(c1.split("#")) & set(c2.split("#")) != set([]):
count += 1
feat_1 = Feature("", 1, {1: count})
#2. POS of main verb
Arg1_MV_POS = dict_util.get_main_verb_pos(relation, "Arg1", parse_dict)
Arg2_MV_POS = dict_util.get_main_verb_pos(relation, "Arg2", parse_dict)
MV_POS_feature_list = Arg1_MV_POS + Arg2_MV_POS
MV_POS_feature = get_feature_by_list(MV_POS_feature_list)
return util.mergeFeatures([feat_1, MV_POS_feature])
def prev_context_conn_sense(relation, parse_dict, implicit_context_dict):
# load dict
dict_prev_context_conn_sense = Non_Explicit_dict().dict_prev_context_conn_sense
# feature
prev_context_conn_sense = dict_util.get_prev_context_conn_sense(relation, parse_dict, implicit_context_dict)
return get_feature_by_feat(dict_prev_context_conn_sense, prev_context_conn_sense)
def arg_first3_conn_pair(relation, parse_dict):
# load dict
dict_arg_first3_conn_pair = Non_Explicit_dict().dict_arg_first3_conn_pair
''' feature '''
arg_first3_conn_pair = dict_util.get_arg_first3_conn_pair(relation, parse_dict)
return get_feature_by_feat(dict_arg_first3_conn_pair, arg_first3_conn_pair)
def cp_production_rules(relation, parse_dict):
''' load dict '''
dict_cp_production_rules = Non_Explicit_dict().cp_production_rules
''' feature '''
cp_production_rules = dict_util.get_cp_production_rules(relation, parse_dict)#["a|b", "b|e"]
return get_feature_by_feat_list(dict_cp_production_rules, cp_production_rules)
def word_pairs(relation, parse_dict):
''' load dict '''
dict_word_pairs = Non_Explicit_dict().dict_word_pairs
''' feature '''
word_pairs = dict_util.get_word_pairs(relation, parse_dict)#["a|b", "b|e"]
return get_feature_by_feat_list(dict_word_pairs, word_pairs)
def get_Arg_brown_cluster(relation, Arg, parse_dict):
Arg_words = get_Arg_Words_List(relation, Arg, parse_dict)
dict_brown_cluster = Non_Explicit_dict().brown_cluster
Arg_brown_cluster = []
for word in Arg_words:
if word in dict_brown_cluster:
Arg_brown_cluster.append(dict_brown_cluster[word])
return Arg_brown_cluster
def get_polarity(word):
polarity = Non_Explicit_dict().polarity
polarity_stem = Non_Explicit_dict().polarity_stem
pol = ""
if word in polarity:
pol = polarity[word]
else:
stem = util.stem_string(word)
if stem in polarity:
pol = polarity[stem]
if stem in polarity_stem:
pol = polarity_stem[stem]
if pol == "":
return []
else:
return pol.split("|")
def get_inquirer_vec(relation, Arg, parse_dict):
inquirer = Non_Explicit_dict().inquirer
inquirer_stem = Non_Explicit_dict().inquirer_stem
list = [0] * 42
verb_tag = ["MD", "VB", "VBD", "VBG", "VBN", "VBP", "VBZ"]
# ["NN/dog" ,"NNS/joks]
word_list = _get_Arg_word_pos_list(relation, Arg, parse_dict)
for item in word_list:
if item == "":
continue
tag, word = item.split("/")[:2]
if tag not in verb_tag:
continue
word = word.lower()
if word in inquirer.keys():
list = _merge(list, inquirer[word])
else:
stem = util.stem_string(word)
if stem in inquirer.keys():
list = _merge(list, inquirer[stem])
elif stem in inquirer_stem.keys():
list = _merge(list, inquirer_stem[stem])
return list
def get_word_MPQA_polarity(word_pos_tuple):
n_stemmed_word_pos_dict = Non_Explicit_dict().n_stemmed_word_pos_dict
y_stemmed_word_pos_dict = Non_Explicit_dict().y_stemmed_word_pos_dict
word, pos = word_pos_tuple
if word_pos_tuple in n_stemmed_word_pos_dict:
return n_stemmed_word_pos_dict[word_pos_tuple]
elif (word, "anypos") in n_stemmed_word_pos_dict:
return n_stemmed_word_pos_dict[(word, "anypos")]
# stem
word = util.stem_string(word)
if (word, pos) in y_stemmed_word_pos_dict:
return y_stemmed_word_pos_dict[(word, pos)]
elif (word, "anypos") in y_stemmed_word_pos_dict:
return y_stemmed_word_pos_dict[(word, "anypos")]
# no match
return ("NULL", "NULL")
def dependency_rules(relation, parse_dict):
''' load dict '''
dict_dependency_rules = Non_Explicit_dict().dict_dependency_rules
''' feature '''
Arg1_dependency_rules = dict_util.get_Arg_dependency_rules(relation, "Arg1", parse_dict)
Arg2_dependency_rules = dict_util.get_Arg_dependency_rules(relation, "Arg2", parse_dict)
Arg1_and_Arg2_dependency_rules = list(set(Arg1_dependency_rules) & set(Arg2_dependency_rules))
feat_Arg1 = get_feature_by_feat_list(dict_dependency_rules, Arg1_dependency_rules)
feat_Arg2 = get_feature_by_feat_list(dict_dependency_rules, Arg2_dependency_rules)
feat_Arg1_and_Arg2 = get_feature_by_feat_list(dict_dependency_rules, Arg1_and_Arg2_dependency_rules)
return util.mergeFeatures([feat_Arg1, feat_Arg2, feat_Arg1_and_Arg2])
def get_word2vec_cluster_pairs(relation, parse_dict):
Arg1_words = get_Arg_Words_List(relation, "Arg1", parse_dict)
Arg2_words = get_Arg_Words_List(relation, "Arg2", parse_dict)
dict_word2vec_cluster = Non_Explicit_dict().word2vec_cluster
word2vec_cluster_pairs = []
for word1 in Arg1_words:
for word2 in Arg2_words:
if word1 in dict_word2vec_cluster and word2 in dict_word2vec_cluster:
word2vec_cluster_pairs.append("%s|%s" %
(dict_word2vec_cluster[word1],
dict_word2vec_cluster[word2]))
return word2vec_cluster_pairs
def production_rules(relation, parse_dict):
'''load dict '''
dict_production_rules = Non_Explicit_dict().dict_production_rules
''' feature '''
Arg1_production_rules = dict_util.get_Arg_production_rules(relation, "Arg1", parse_dict)
Arg2_production_rules = dict_util.get_Arg_production_rules(relation, "Arg2", parse_dict)
Arg1_and_Arg2_production_rules = list(set(Arg1_production_rules) & set(Arg2_production_rules))
Arg1_production_rules = ["Arg1_%s" % rule for rule in Arg1_production_rules]
Arg2_production_rules = ["Arg2_%s" % rule for rule in Arg2_production_rules]
Both_production_rules = ["Both_%s" % rule for rule in Arg1_and_Arg2_production_rules]
rules = Arg1_production_rules + Arg2_production_rules + Both_production_rules
return get_feature_by_feat_list(dict_production_rules, rules)
def arg_brown_cluster(relation, parse_dict):
# load dict
dict_brown_cluster = Non_Explicit_dict().dict_Arg_brown_cluster
''' feature '''
Arg1_brown_cluster = dict_util.get_Arg_brown_cluster(relation, "Arg1", parse_dict)
Arg2_brown_cluster = dict_util.get_Arg_brown_cluster(relation, "Arg2", parse_dict)
Both_brown_cluster = list(set(Arg1_brown_cluster) & set(Arg2_brown_cluster))
Arg1_only = list(set(Arg1_brown_cluster) - set(Arg2_brown_cluster))
Arg2_only = list(set(Arg2_brown_cluster) - set(Arg1_brown_cluster))
Arg1_brown_cluster = ["Arg1_%s" % x for x in Arg1_only]
Arg2_brown_cluster = ["Arg2_%s" % x for x in Arg2_only]
Both_brown_cluster = ["Both_%s" % x for x in Both_brown_cluster]
cluster = Arg1_brown_cluster + Arg2_brown_cluster + Both_brown_cluster
return get_feature_by_feat_list(dict_brown_cluster, cluster)
def is_negate_MPQA(index, word_list):
negate_words = Non_Explicit_dict().negate
prev1 = "NULL"
prev2 = "NULL"
prev3 = "NULL"
if index - 1 >= 0:
prev1 = word_list[index - 1]
if index - 2 >= 0:
prev2 = word_list[index - 2]
if index - 3 >= 0:
prev3 = word_list[index - 3]
prev_words = [prev1, prev2, prev3]
if set(prev_words) & set(negate_words) != set([]):
return True
else:
return False
def Arg_word2vec(relation, parse_dict):
''' load dict '''
dict_word2vec = Non_Explicit_dict().word2vec_dict
''' feature '''
Arg1_words = dict_util._get_lower_case_lemma_words(relation, "Arg1",
parse_dict)
Arg2_words = dict_util._get_lower_case_lemma_words(relation, "Arg2",
parse_dict)
Arg1_words = list(set(Arg1_words))
Arg2_words = list(set(Arg2_words))
Arg1_vec = [0.0] * 300
Arg1_length = 0
for word in Arg1_words:
if word in dict_word2vec:
vec = dict_word2vec[word]
Arg1_vec = util.vec_plus_vec(Arg1_vec, vec)
Arg1_length += 1
Arg2_vec = [0.0] * 300
Arg2_length = 0
for word in Arg2_words:
if word in dict_word2vec:
vec = dict_word2vec[word]
Arg2_vec = util.vec_plus_vec(Arg2_vec, vec)
Arg2_length += 1
# average
if Arg1_length != 0:
Arg1_vec = [v / Arg1_length for v in Arg1_vec]
if Arg2_length != 0:
Arg2_vec = [v / Arg2_length for v in Arg2_vec]
feat1 = get_feature_by_list(Arg1_vec)
feat2 = get_feature_by_list(Arg2_vec)
return util.mergeFeatures([feat1, feat2])
