95 lines
3.1 KiB
Python
95 lines
3.1 KiB
Python
|
import pandas as pd
|
||
|
|
import numpy as np
|
||
|
|
import torch
|
||
|
|
from nltk.tokenize import word_tokenize
|
||
|
|
import gensim.downloader
|
||
|
|
|
||
|
|
|
||
|
|
x_train = pd.read_table('train/in.tsv', sep='\t', error_bad_lines=False, quoting=3, header=None, names=['content', 'id'], usecols=['content'])
|
||
|
|
y_train = pd.read_table('train/expected.tsv', sep='\t', error_bad_lines=False, quoting=3, header=None, names=['label'])
|
||
|
|
x_dev = pd.read_table('dev-0/in.tsv', sep='\t', error_bad_lines=False, header=None, quoting=3, names=['content', 'id'], usecols=['content'])
|
||
|
|
x_test = pd.read_table('test-A/in.tsv', sep='\t', error_bad_lines=False, header=None, quoting=3, names=['content', 'id'], usecols=['content'])
|
||
|
|
|
||
|
|
x_train = x_train.content.str.lower()
|
||
|
|
x_dev = x_dev.content.str.lower()
|
||
|
|
x_test = x_test.content.str.lower()
|
||
|
|
|
||
|
|
x_train = [word_tokenize(content) for content in x_train]
|
||
|
|
x_dev = [word_tokenize(content) for content in x_dev]
|
||
|
|
x_test = [word_tokenize(content) for content in x_test]
|
||
|
|
|
||
|
|
word2vec = gensim.downloader.load("word2vec-google-news-300")
|
||
|
|
|
||
|
|
def document_vector(doc):
|
||
|
|
"""Create document vectors by averaging word vectors. Remove out-of-vocabulary words."""
|
||
|
|
return np.mean([word2vec[w] for w in doc if w in word2vec] or [np.zeros(300)], axis=0)
|
||
|
|
|
||
|
|
x_train = [document_vector(doc) for doc in x_train]
|
||
|
|
x_dev = [document_vector(doc) for doc in x_dev]
|
||
|
|
x_test = [document_vector(doc) for doc in x_test]
|
||
|
|
|
||
|
|
class NeuralNetwork(torch.nn.Module):
|
||
|
|
def __init__(self, hidden_size):
|
||
|
|
super(NeuralNetwork, self).__init__()
|
||
|
|
self.l1 = torch.nn.Linear(300, hidden_size)
|
||
|
|
self.l2 = torch.nn.Linear(hidden_size, 1)
|
||
|
|
|
||
|
|
def forward(self, x):
|
||
|
|
x = self.l1(x)
|
||
|
|
x = torch.relu(x)
|
||
|
|
x = self.l2(x)
|
||
|
|
x = torch.sigmoid(x)
|
||
|
|
return x
|
||
|
|
|
||
|
|
|
||
|
|
hidden_size = 600
|
||
|
|
epochs = 5
|
||
|
|
batch_size = 15
|
||
|
|
model = NeuralNetwork(hidden_size)
|
||
|
|
criterion = torch.nn.BCELoss()
|
||
|
|
optimizer = torch.optim.SGD(model.parameters(), lr=0.01)
|
||
|
|
|
||
|
|
for epoch in range(epochs):
|
||
|
|
model.train()
|
||
|
|
for i in range(0, y_train.shape[0], batch_size):
|
||
|
|
X = x_train[i:i+batch_size]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
y = y_train[i:i+batch_size]
|
||
|
|
y = torch.tensor(y.astype(np.float32).to_numpy()).reshape(-1, 1)
|
||
|
|
|
||
|
|
outputs = model(X.float())
|
||
|
|
loss = criterion(outputs, y)
|
||
|
|
|
||
|
|
optimizer.zero_grad()
|
||
|
|
loss.backward()
|
||
|
|
optimizer.step()
|
||
|
|
|
||
|
|
y_dev = []
|
||
|
|
y_test = []
|
||
|
|
|
||
|
|
model.eval()
|
||
|
|
|
||
|
|
with torch.no_grad():
|
||
|
|
for i in range(0, len(x_dev), batch_size):
|
||
|
|
X = x_dev[i:i+batch_size]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
outputs = model(X.float())
|
||
|
|
prediction = (outputs > 0.5)
|
||
|
|
y_dev.extend(prediction)
|
||
|
|
|
||
|
|
for i in range(0, len(x_test), batch_size):
|
||
|
|
X = x_test[i:i+batch_size]
|
||
|
|
X = torch.tensor(X)
|
||
|
|
outputs = model(X.float())
|
||
|
|
y = (outputs > 0.5)
|
||
|
|
y_test.extend(prediction)
|
||
|
|
|
||
|
|
|
||
|
|
y_dev = np.asarray(y_dev, dtype=np.int32)
|
||
|
|
y_test = np.asarray(y_test, dtype=np.int32)
|
||
|
|
|
||
|
|
y_dev = pd.DataFrame({'label':y_dev})
|
||
|
|
y_test = pd.DataFrame({'label':y_test})
|
||
|
|
|
||
|
|
y_dev.to_csv(r'dev-0/out.tsv', sep='\t', index=False, header=False)
|
||
|
|
y_test.to_csv(r'test-A/out.tsv', sep='\t', index=False, header=False)
|