paranormal-or-skeptic/run.ipynb

import numpy as np
import gensim
import torch
import pandas as pd
from gensim.models import Word2Vec
from gensim import downloader
from sklearn.feature_extraction.text import TfidfVectorizer

BATCH_SIZE = 10
EPOCHS = 100
FEAUTERES = 200

class NeuralNetworkModel(torch.nn.Module):
    
    def __init__(self):
        super(NeuralNetworkModel, self).__init__()
        self.fc1 = torch.nn.Linear(FEAUTERES,500)
        self.fc2 = torch.nn.Linear(500,1)

    def forward(self, x):
        x = self.fc1(x)
        x = torch.relu(x)
        x = self.fc2(x)
        x = torch.sigmoid(x)
        return x

word2vec = downloader.load("glove-twitter-200")

def readData(fileName):  
    with open(f'{fileName}/in.tsv', 'r', encoding='utf8') as f:
        X =  np.array([x.strip().lower() for x in f.readlines()])
    with open(f'{fileName}/expected.tsv', 'r', encoding='utf8') as f:
        y = np.array([int(x.strip()) for x in f.readlines()])
    return X,y

X_file,y_file = readData('dev-0')

x_train_w2v = [np.mean([word2vec[word.lower()] for word in doc.split() if word.lower() in word2vec]
                   or [np.zeros(FEAUTERES)], axis=0) for doc in X_file]

def train_model(X_file,y_file):
    model = NeuralNetworkModel()

    criterion = torch.nn.BCELoss()
    optimizer = torch.optim.ASGD(model.parameters(), lr=0.05)
    for epoch in range(EPOCHS):
        print(epoch)
        loss_score = 0
        acc_score = 0
        items_total = 0
        for i in range(0, y_file.shape[0], BATCH_SIZE):
            x = X_file[i:i+BATCH_SIZE]
            x = torch.tensor(np.array(x).astype(np.float32))
            y = y_file[i:i+BATCH_SIZE]
            y = torch.tensor(y.astype(np.float32)).reshape(-1, 1)
            y_pred = model(x)
            acc_score += torch.sum((y_pred > 0.5) == y).item()
            items_total += y.shape[0]

            optimizer.zero_grad()
            loss = criterion(y_pred, y)
            loss.backward()
            optimizer.step()

            loss_score += loss.item() * y.shape[0]
        
        print((loss_score / items_total), (acc_score / items_total))
    return model

def predict(model,x_file):
    y_dev = []
    with torch.no_grad():
        for i in range(0, len(x_file), BATCH_SIZE):
            x = x_file[i:i+BATCH_SIZE]
            x = torch.tensor(np.array(x).astype(np.float32))
            outputs = model(x)
            y = (outputs > 0.5)
            y_dev.extend(y)
    return y_dev
        

def wrtieToFile(fileName,y_file):
    y_out = []
    for y in y_file:
        y_out.append(int(str(y[0]).split('(')[1].split(')')[0]=='True'))
    with open(f'{fileName}/out.tsv','w',encoding='utf8') as f:
        for y in y_out:
            f.write(f'{y}\n')

model = train_model(x_train_w2v,y_file)

0
0.6414709375416563 0.6464339908952959
1
0.6118579905971953 0.6589529590288316
2
0.5930351529140393 0.677731411229135
3
0.5807589731138194 0.6936646433990895
4
0.5711128521026628 0.7031487101669196
5
0.5637358135638451 0.7065629742033384
6
0.5573145605239321 0.710546282245827
7
0.5521481898931252 0.715288315629742
8
0.5475104518053836 0.7181335356600911
9
0.5430893454028008 0.7202200303490136
10
0.5395108298066443 0.7236342943854325
11
0.5361589408495177 0.7257207890743551
12
0.53314527610885 0.7270485584218513
13
0.5298747769267226 0.7297040971168437
14
0.5269876997833096 0.7319802731411229
15
0.5245049590914763 0.7336874051593323
16
0.5220209190930057 0.7363429438543247
17
0.5203242429527871 0.7365326251896813
18
0.5182899421417297 0.737670713201821
19
0.5155506848000069 0.7401365705614568
20
0.5131794015095429 0.7403262518968133
21
0.5113656374375719 0.7412746585735963
22
0.5092821710139558 0.7420333839150227
23
0.5067137854063547 0.7441198786039454
24
0.5047900934558085 0.745257966616085
25
0.5025694217866397 0.7488619119878603
26
0.5007175219885451 0.7486722306525038
27
0.4981631609315847 0.747154779969651
28
0.4961598192105615 0.7498103186646434
29
0.49438970515077685 0.7507587253414264
30
0.49240998727621366 0.7507587253414264
31
0.4907134136267018 0.7520864946889226
32
0.48826086573438415 0.7541729893778453
33
0.4871711270185541 0.7560698027314112
34
0.48422483688330614 0.756638846737481
35
0.48217912709371546 0.7604324734446131
36
0.48009182657475535 0.761380880121396
37
0.4778907883217013 0.7632776934749621
38
0.47551582766660067 0.7621396054628224
39
0.47324845619635353 0.7646054628224582
40
0.47138607904755925 0.7653641881638846
41
0.4684638544374424 0.7659332321699545
42
0.4662012148575012 0.7685887708649469
43
0.46414706633568986 0.7693474962063733
44
0.4620490156040613 0.7702959028831563
45
0.46027336999977486 0.7706752655538694
46
0.4574687189093264 0.7746585735963581
47
0.45456105805311653 0.7748482549317147
48
0.45308226045385025 0.7769347496206374
49
0.44969080617490237 0.7792109256449166
50
0.4477136310823092 0.77902124430956
51
0.44523295281067693 0.7841426403641881
52
0.44300158465442235 0.7835735963581184
53
0.44147631555388656 0.7852807283763278
54
0.43824701448718767 0.78850531107739
55
0.437326367692923 0.7936267071320182
56
0.43404240863558824 0.7936267071320182
57
0.43146262304328825 0.7959028831562974
58
0.429094969041996 0.7938163884673748
59
0.42631421763059857 0.7977996965098634
60
0.4239590879280985 0.798937784522003
61
0.4216488930983229 0.8014036418816388
62
0.41922062316595693 0.8033004552352049
63
0.417561381201688 0.8053869499241275
64
0.4144452941633637 0.8051972685887708
65
0.41305530049212064 0.8080424886191199
66
0.410686616688311 0.8072837632776935
67
0.4076426998430889 0.8114567526555387
68
0.4061218895193342 0.811267071320182
69
0.4029337710281198 0.8139226100151745
70
0.40099998707395496 0.8143019726858877
71
0.39854915830701004 0.8133535660091047
72
0.39473064304845285 0.8201820940819423
73
0.3931978788616896 0.8198027314112292
74
0.3905544553760422 0.8218892261001517
75
0.3894510168316513 0.8211305007587253
76
0.38586248252229916 0.8247344461305007
77
0.3851398667786977 0.8256828528072838
78
0.38457902678046857 0.8247344461305007
79
0.3803209278461197 0.8272003034901366
80
0.37845283393127693 0.8287177541729894
81
0.37618811287505943 0.8294764795144158
82
0.37400476449368486 0.8323216995447648
83
0.3726042910890261 0.8332701062215478
84
0.36963997851373215 0.8338391502276176
85
0.3680792153446917 0.8363050075872535
86
0.36542417398160704 0.8361153262518968
87
0.36405448698366627 0.8376327769347496
88
0.3595154614517061 0.8423748103186647
89
0.35860147739566967 0.8419954476479514
90
0.3578952589836848 0.8404779969650986
91
0.35602253879814755 0.8414264036418816
92
0.3523210818386087 0.8446509863429439
93
0.34952340000598764 0.8480652503793626
94
0.3513405356550524 0.845030349013657
95
0.349314306160274 0.8493930197268589
96
0.34516190266595626 0.8492033383915023
97
0.34279035948137776 0.8524279210925645
98
0.34358633996576793 0.8518588770864947
99
0.3396215371445545 0.8528072837632777

y_dev=predict(model,x_train_w2v)

wrtieToFile("dev-0",y_dev)

with open(f'test-A/in.tsv', 'r', encoding='utf8') as f:
    X =  np.array([x.strip().lower() for x in f.readlines()])

x_train_w2v = [np.mean([word2vec[word.lower()] for word in doc.split() if word.lower() in word2vec]
                   or [np.zeros(FEAUTERES)], axis=0) for doc in X]

y_dev=predict(model,x_train_w2v)

wrtieToFile("test-A",y_dev)