-
Notifications
You must be signed in to change notification settings - Fork 0
/
run_experiment_curated_cases.py
226 lines (193 loc) · 9.96 KB
/
run_experiment_curated_cases.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
import os
import pandas as pd
import numpy as np
from tasks import wsd
from pathlib import Path
#from utils import nlp_tools
#from parhugin import multiFunc
from gensim.models import Word2Vec
from utils.classificaton_utils import binarize#,generate_definition_df, vectorize_target_expressions
from tqdm import tqdm
from sklearn.svm import LinearSVC
from sklearn.linear_model import Perceptron
from sklearn.neural_network import MLPClassifier
WEMB_MODEL_PATH = "models/w2v_004/w2v_words.model"
def eval_sense(lemma,
pos,
senses,
start,
end,
train_on_dev,
eval_mode,
relations,
vector_cols,
filter_val,
filter_test,
wemb_model):
print(f'STARTING AT {start}; ENDING AT {end}')
df_train, df_val, df_test = binarize(lemma=lemma,
pos=pos,
senses=senses,
start=start,
end=end,
relations=relations,
eval_mode=eval_mode,
filter_val_by_year=filter_val,
filter_test_by_year=filter_test,
strict_filter=True)
# no quotations for sense and timeframe
if df_train is None:
return None
if train_on_dev:
df_train = pd.concat([df_train, df_val], axis=0)
df_train.reset_index(inplace=True,drop=True)
for vector_col in vector_cols:
senseid2label = dict(df_test[['sense_id','label']].values)
df_test[f"bert_centroid_sense_{vector_col}"] = df_test.apply(wsd.bert_sense_centroid_vector,
senseid2label= senseid2label,
vector_col=vector_col,
df_train = df_train, axis=1)
#df_test[f"bert_ts_nearest_centroid_sense_{vector_col}"] = df_test.apply(wsd.bert_ts_sense_centroid_vector,
# senseid2label= senseid2label,
# ts_method='nearest',
# vector_col=vector_col,
# df_train = df_train, axis=1)
df_test[f"bert_ts_weighted_centroid_sense_{vector_col}"] = df_test.apply(wsd.bert_ts_sense_centroid_vector,
senseid2label= senseid2label,
ts_method='weighted',
vector_col=vector_col,
df_train = df_train, axis=1)
# TO DO: uncomment this after merging with dev
#df_test[f"bert_ts_weighted_centroid_sense_{vector_col}"] = df_test.apply(wsd.bert_ts_sense_centroid_vector,
# senseid2label= senseid2label,
# ts_method='weighted_past',
# vector_col=vector_col,
# df_train = df_train, axis=1)
print(f'[LOG] traing classifier for {senses} [BERT model = {vector_col}]' )
X,y = list(df_train[vector_col].values), list(df_train.label.values)
#svm_model = LinearSVC(random_state=0, C=.1, tol=1e-5,class_weight='balanced')
#svm_model.fit(X,y)
#df_test[f"bert_svm_{vector_col}"] = wsd.clf_svm(vector_col,df_test, svm_model)
#perc_model = Perceptron(validation_fraction=.2, early_stopping=True,class_weight='balanced')
#perc_model.fit(X,y)
#df_test[f"bert_perceptron_{vector_col}"] = wsd.clf_perceptron(vector_col,df_test, perc_model)
mlperc_model = MLPClassifier(validation_fraction=.2, early_stopping=True, solver='lbfgs',activation='relu')
mlperc_model.fit(X,y)
df_test[f"bert_ml_perceptron_{vector_col}"] = wsd.clf_perceptron(vector_col,df_test, mlperc_model)
return df_test
def run(lemma,
pos,
senses,
start,
end,
vector_cols,
eval_mode,
relations,
train_on_dev,
wemb_model,
filter_val,
filter_test,
results_filename,
results_path_base):
df_test = eval_sense(lemma=lemma,
pos=pos,
senses=senses,
start=start,
end=end,
train_on_dev=train_on_dev,
eval_mode=eval_mode,
relations=relations,
vector_cols=vector_cols,
filter_val=filter_val,
filter_test=filter_test,
wemb_model=wemb_model)
results_path = os.path.join(results_path_base, f"{lemma}_{pos}", eval_mode)
results_filename = results_filename + "+".join(sorted(relations)) + ".csv"
Path(results_path).mkdir(parents=True, exist_ok=True)
# IF df_test is None, create an empty DataFrame
if not isinstance(df_test, type(None)):
baselines = ["id_x",'label','year','quotation_id'] #'random','def_tok_overlap_ranking', 'sent_embedding', 'w2v_lesk_ranking'] # ,'svm_wemb_baseline'
bert_methods = [[f"bert_centroid_sense_{vector_col}",
#f"bert_ts_nearest_centroid_sense_{vector_col}",
f"bert_ts_weighted_centroid_sense_{vector_col}",
f"bert_ml_perceptron_{vector_col}"
]
for vector_col in vector_cols]
bert_methods = [i for tm in bert_methods for i in tm]
out_df = df_test.filter(baselines + bert_methods, axis=1)
else:
out_df = pd.DataFrame()
out_df.to_csv(os.path.join(results_path, results_filename), index=False)
def run_experiment(direction='vertical'):
if direction == 'vertical':
RELATIONS = ['seed'] # 'synonym'
EVAL_MODE = 'lemma' #'lemma_etal'
RESULTS_PATH_BASE = 'results_curated_1920_seed'
elif direction == 'horizontal':
RELATIONS = ['seed','synonym'] # ''
EVAL_MODE = 'lemma_etal' #'lemma_etal'
RESULTS_PATH_BASE = 'results_curated_1920_syn'
WEMB_MODEL = Word2Vec.load(WEMB_MODEL_PATH)
TRAIN_ON_DEV = True
# argument the change by experiment change
FILTER_VAL = True
FILTER_TEST = True
VECTOR_COLS = ['vector_bert_base_-1,-2,-3,-4_mean',
"vector_blert_base_-1,-2,-3,-4_mean",
'vector_bert_1850_-1,-2,-3,-4_mean'
]
START = 1760
END = 1920
FILTER_VAL = False
FILTER_TEST = True
#RESULTS_PATH_BASE = 'results_curated_1920_seed'
words = {
("slave_sense_1_ethnicity",'slave','NN'): {"slave_nn01-22495496", "slave_nn01-22495604","slave_nn02-22498415"},
("slave_sense_2_humansubmission",'slave','NN'): {"slave_nn01-22495881","slave_nn01-22496175","slave_nn01-22496245","slave_nn01-22496365","slave_nn01-22498301"},
("slave_sense_3_object",'slave','NN'): {"slave_nn01-22496432","slave_nn01-22496462"},
("power_sense_1_ability",'power','NN'): {"power_nn01-28685348","power_nn01-28684730","power_nn01-28684973","power_nn01-28685458",
"power_nn01-28685148","power_nn01-28686226","power_nn01-28684355","power_nn01-28685727","power_nn01-110297806","power_nn01-28684538"},
("power_sense_2_legal",'power','NN'): {"power_nn01-28685686","power_nn01-28685253","power_nn01-28688384","power_nn01-28687020",
"power_nn01-28684910","power_nn01-110297806","power_nn01-28686314","power_nn01-28686446","power_nn01-28686005","power_nn01-28686566"},
("power_sense_3_technical","power","NN"): {"power_nn01-28688328","power_nn01-52944172","power_nn01-28685989","power_nn01-224965906",
"power_nn01-28687020","power_nn01-28687898","power_nn01-52944158","power_nn01-28688218","power_nn01-28688482"},
("power_sense_4_maths","power","NN"): {"power_nn01-28687845","power_nn01-28687774","power_nn01-28687471","power_nn01-28687446",
"power_nn01-28687561","power_nn01-28687732"},
("labour_sense_1_physical","labour","NN"): {"labour_nn01-39839017","labour_nn01-39838984","labour_nn01-39839322","labour_nn01-39838736",
"labour_nn01-185646201","labour_nn01-185646093","labour_nn01-39839548","labour_nn01-39839598"},
("labour_sense_2_human","labour","NN"): {"labour_nn01-184772185","labour_nn01-184772185","labour_nn01-39839092"},
("labour_sense_3_figurative","labour","NN"): {"labour_nn01-39839738","labour_nn01-39839676"},
("labour_sense_4_political",'labour',"NN"): {"labour_nn01-39839157","labour_nn01-39839281"},
("machine_non_figurative","machine","NN"): {"machine_nn01-38473945","machine_nn01-38474233","machine_nn01-38474301","machine_nn01-38474548",
"machine_nn01-38475099","machine_nn01-38475046","machine_nn01-38475013","machine_nn01-38474974","machine_nn01-38474877",
"machine_nn01-38475494","machine_nn01-38474820","machine_nn01-38475164","machine_nn01-38475923","machine_nn01-38475286","machine_nn01-38474607"},
("machine_figurative","machine","NN"): {"machine_nn01-38474140","machine_nn01-38474405","machine_nn01-38475994","machine_nn01-38476096",
"machine_nn01-38476316","machine_nn01-38476397","machine_nn01-38476566","machine_nn01-38476245","machine_nn01-38475835"}
}
errors = []
for name, senses in words.items():
name,lemma,pos = name
try:
run(lemma,
pos,
senses,
start=START,
end=END,
vector_cols=VECTOR_COLS,
eval_mode=EVAL_MODE,
relations=RELATIONS,
train_on_dev=TRAIN_ON_DEV,
wemb_model=WEMB_MODEL,
filter_val=FILTER_VAL,
filter_test=FILTER_TEST,
results_filename=name,
results_path_base=RESULTS_PATH_BASE)
except Exception as e:
print(name,e)
errors.append(name)
print("Done.")
print("Errors with the following senses:")
print(errors)
if __name__=="__main__":
run_experiment(direction='vertical')
run_experiment(direction='horizontal')