-
Notifications
You must be signed in to change notification settings - Fork 0
/
evaluate_result.py
199 lines (164 loc) · 7.97 KB
/
evaluate_result.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
import json
import logging
import os
import sys
from dataclasses import dataclass, field
from typing import Optional
from datetime import datetime
import nltk # Here to have a nice missing dependency error message early on
import numpy as np
import pandas as pd
from datasets import load_dataset, load_metric
from unexpected_mask import unexpected_sent_score, get_diff_score
from hypo_select import get_hypo_pair_score, get_hypo_judge_score
from filelock import FileLock
from bert_score import BERTScorer
from nltk.translate.bleu_score import corpus_bleu
from nltk import word_tokenize
#from hypo_select import get_hypo_pair_score, get_hypo_judge_score
def get_length(s):
return len(word_tokenize(s))
def compute_metrics(baseline, test_paras, cands):
assert len(test_paras) == len(cands)
cands = [[c] for c in cands]
para_cand_map = dict(zip(test_paras, cands))
df = pd.read_csv("HYPO.tsv", delimiter='\t',
header=None, names=['hypo', 'para', 'mnc'])
df = df.dropna(subset=['hypo', 'para'])
para = df.para.values.tolist()
para = [s.strip() for s in para]
refs = df.hypo.values.tolist()
refs = [s.strip() for s in refs]
# assert len(refs) == len(list(para_cand_map.keys()))
para_ref_map = dict(zip(para,refs))
para_hypo = []
scorer = BERTScorer(lang="en", rescale_with_baseline=True, device='cpu')
df_cand_scores = pd.DataFrame(columns=['para','cand', 'hypo_score', 'pair_score', 'unexpected_score', 'select_score'])
refs = []
gens = []
best_cand_scores = []
repetition_idx = []
for idx, para in enumerate(para_cand_map.keys()):
cands = para_cand_map[para]
unexpected_scores = np.array([unexpected_sent_score(s) for s in cands])
pair_scores = get_hypo_pair_score(para, cands)
hypo_scores = get_hypo_judge_score(cands)
select_scores = pair_scores + hypo_scores
# select_scores = pair_scores + hypo_scores + unexpected_scores
# select_scores = (hypo_scores>0.5)*pair_scores*hypo_scores
# select_scores = pair_scores
cand_scores = []
for i in range(len(cands)):
cand_scores.append([para, cands[i], hypo_scores[i], pair_scores[i], unexpected_scores[i], select_scores[i]])
df_tmp = pd.DataFrame(cand_scores, columns=['para','cand', 'hypo_score', 'pair_score', 'unexpected_score','select_score'])
df_tmp = df_tmp.sort_values(by=['select_score', 'hypo_score', 'unexpected_score','pair_score'], ascending=[False, False, False, False])
df_cand_scores = df_cand_scores.append(df_tmp)
if not select_scores.any(): # all zeros
max_idx = np.argmax(hypo_scores)
best_cand = cands[max_idx]
else:
max_idx = np.argmax(select_scores)
best_cand = cands[max_idx]
"""
if idx in [2, 4, 7, 9, 10, 11, 12, 15, 16, 17, 20, 22, 24, 25, 27, 28, 29, 30, 32, 34, 37, 38, 41, 45, 52, 53, 54, 55, 57, 58, 60, 62, 64, 66, 67, 69]:
# print(para, best_cand)
continue
"""
if best_cand == para:
repetition_idx.append(idx)
# continue
diff_score = get_diff_score(para, best_cand)
length = get_length(best_cand)
best_cand_scores.append([hypo_scores[max_idx], unexpected_scores[max_idx], pair_scores[max_idx], select_scores[max_idx], diff_score, length])
para_hypo.append([para,best_cand])
gens.append(best_cand)
refs.append(para_ref_map[para])
print("copy:", len(repetition_idx), "total:", len(refs))
P, R, F1 = scorer.score(gens, refs)
bert_score = F1.tolist()
metrics = dict()
#max_test_samples = data_args.max_test_samples if data_args.max_test_samples is not None else len(test_dataset)
#metrics["test_samples"] = min(max_test_samples, len(test_dataset))
metrics["BERTScore"] = sum(bert_score)/len(bert_score)
r = [[row.split()] for row in refs]
c = [row.split() for row in gens]
metrics["BLEU"] = corpus_bleu(r, c)*100
metrics["BLEU-1"] = corpus_bleu(r, c,weights=(1, 0, 0, 0))*100
metrics["BLEU-2"] = corpus_bleu(r, c,weights=(0, 1, 0, 0))*100
metrics["BLEU-3"] = corpus_bleu(r, c,weights=(0, 0, 1, 0))*100
metrics["BLEU-4"] = corpus_bleu(r, c,weights=(0, 0, 0, 1))*100
best_cand_scores = np.array(best_cand_scores)
best_cand_scores_avg = np.average(best_cand_scores, axis=0)
metrics['hypo_score'] = best_cand_scores_avg[0]
metrics['unexpected_score'] = best_cand_scores_avg[1]
metrics['pair_score'] = best_cand_scores_avg[2]
metrics['select_score'] = best_cand_scores_avg[3]
metrics["difference"] = best_cand_scores_avg[4]
metrics["length"] = best_cand_scores_avg[5]
# metrics["EM"] = sum([gen == para for gen, para in zip(gens, test_paras)]) / len(gens)
metrics["EM"] = sum([item[0] == item[1] for item in para_hypo]) / len(gens)
print("*"*20, baseline, "*"*20)
print(metrics)
df = pd.read_csv("bart_hypo_hypo_test.csv", header=None, names=['para', 'cand'])
test_paras = df.para.values.tolist()
#test_cands = df.cand.values.tolist()
#compute_metrics("BART-HYPO", test_paras, test_cands)
output_file = "HYPO_test_for_eval.csv"
df = pd.read_csv(output_file)
mover = df.MOVER.values.tolist()
R1 = df.R1.values.tolist()
R3 = df.R3.values.tolist()
BART = df.BART.values.tolist()
BART_PARA = df.BART_PARA.values.tolist()
HUMAN = df.HUMAN.values.tolist()
NAIVE = test_paras
df = pd.read_csv("model_save/bart_base_/checkpoint-10479/test_final_output.csv")
MOVER = df['hyperbolic sentence'].tolist()
df = pd.read_csv("model_save/bart_hypo_hypo_test/test_final_output.csv", header=None, names=['literal sentence', 'hyperbolic sentence'])
BART_HYPO = df['hyperbolic sentence'].tolist()
df = pd.read_csv("model_save/mover_test_no_tag/test_final_output.csv")
MOVER_HYPO_SELECT = df['hyperbolic sentence'].tolist()
df = pd.read_csv("model_save/mover_hypo_train/test_final_output.csv")
MOVER_HYPO_TRAIN = df['hyperbolic sentence'].tolist()
df = pd.read_csv("model_save/mover_hypo_train_hypo-xl/test_final_output.csv")
MOVER_HYPO_TRAIN_HYPO_XL = df['hyperbolic sentence'].tolist()
"""
with open("model_save/debug_bart_hypo_whole_input/test_generations.txt", 'r') as inF:
BART_HYPO_REAL = [l.strip() for l in inF.readlines()]
assert NAIVE != BART_HYPO_REAL # asserted!
"""
with open("model_save/mover_bart_hypo/test_generations.txt", 'r') as inF:
MOVER_BART_HYPO = [l.strip() for l in inF.readlines()]
assert NAIVE != MOVER_BART_HYPO
with open("output/paraphrase_generations.txt", 'r') as inF:
PARA = [l.strip() for l in inF.readlines()]
assert NAIVE != PARA
with open("model_save/debug_bart_base_para_paws_qqp_hypo/test_generations.txt", 'r') as inF:
BART_PARA_HYPO = [l.strip() for l in inF.readlines()]
assert NAIVE != BART_PARA_HYPO
with open("model_save/debug_bart_base_para_paws_qqp_whole_input/test_generations.txt", 'r') as inF:
BART_PARA_PAWS_QQP = [l.strip() for l in inF.readlines()]
assert NAIVE != BART_PARA_PAWS_QQP
#compute_metrics("BART_HYPO_REAL", test_paras, BART_HYPO_REAL)
compute_metrics("MOVER", test_paras, MOVER)
compute_metrics("R1", test_paras, R1)
compute_metrics("R3", test_paras, R3)
compute_metrics("BART", test_paras, BART)
#compute_metrics("BART-PARA", test_paras, BART_PARA)
compute_metrics("HUMAN", test_paras, HUMAN)
compute_metrics("NAIVE", test_paras, NAIVE)
#compute_metrics("BART_HYPO", test_paras, BART_HYPO)
#compute_metrics("MOVER_BART_HYPO", test_paras, MOVER_BART_HYPO)
#compute_metrics("MOVER_HYPO_SELECT", test_paras, MOVER_HYPO_SELECT)
#compute_metrics("MOVER_HYPO_TRAIN", test_paras, MOVER_HYPO_TRAIN)
#compute_metrics("MOVER_HYPO_TRAIN_HYPO_XL", test_paras, MOVER_HYPO_TRAIN_HYPO_XL)
#compute_metrics("PARA", test_paras, PARA)
#compute_metrics("BART_PARA_HYPO", test_paras, BART_PARA_HYPO)
#compute_metrics("BART_PARA_PAWS_QQP", test_paras, BART_PARA_PAWS_QQP)
# test_paras = df.para.values.tolist()
"""
fIn=open("bart_para_outputs.txt", 'r')
cands =[[l.strip()] for l in fIn.readlines()]
"""
# cands = [[p.strip()] for p in test_paras]
# cands = df.cand.values.tolist()