/Chinese_News_Classification

Primary LanguagePython

Chinese News Classification

使用 pytorch 練習 GRU 進行新聞分類

Enviroment

  • OS: unbuntu 18.04 LTS
  • GPU: GTX 1070 8G
  • RAM: 16G

Requiremnet

  • pytorch 1.0
  • jieba
  • numpy
  • scikit-learn

Dataset

10 label names:

体育, 财经, 房产, 家居, 教育, 科技, 时尚, 时政, 游戏, 娱乐

link

  • train dataset: 5000 * 10
  • test dataset: 1000 * 10
  • valid dataset: 500 * 10

word embedding

Mixed-large 综合 in link

word size : about 1292608

Model

Model(
  (embedding_layer): Embedding(1291384, 300)
  (bigru): GRU(300, 64, num_layers=2, dropout=0.2, bidirectional=True)
  (output_layer): Linear(in_features=256, out_features=512, bias=True)
)

Result