-
Notifications
You must be signed in to change notification settings - Fork 0
/
fix_known_documents.py
46 lines (37 loc) · 1.25 KB
/
fix_known_documents.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
#!/usr/bin/env python3
"""
Set labels of the test documents that appear in the training set
Requires around 4 Gb memory
"""
import csv
import numpy as np
import pandas
from train_kernel_svm import load_test_voc_and_bows, intersection_kernel, dataset2bows, normalize_it
df = pandas.read_pickle('data/train2.p', compression='gzip')
train_data = list(map(str, df['content']))
train_labels = list(map(int, df['label']))
test_labels = []
with open("submission26.5.csv", newline='') as f:
f.readline()
reader = csv.reader(f)
for row in reader:
test_labels.append(int(row[0]))
voc, test_bows = load_test_voc_and_bows()
print("Normalizing test BOWs")
normalize_it(test_bows)
print("Computing training BOWs")
_, train_bows = dataset2bows(train_data, voc)
print("Normalizing training BOWs")
normalize_it(train_bows)
print("Computing kernel. Please, wait around 10 min")
K = intersection_kernel(test_bows, train_bows)
count = 0
for i, row in enumerate(K):
if np.max(row) > 0.95:
test_labels[i] = train_labels[np.argmax(row)]
count += 1
print(count, "/", len(test_labels))
with open("submission26.5_patched.csv", 'w') as fw:
fw.write("topic,index\n")
for i, y in enumerate(test_labels):
fw.write(f"{y},{i}\n")