TextCNN by using Pytorch1.0 python3.6 + torchtext0.4.0
语料采用清华的新闻语料(精简后10个类别,每个类别5000样本)
语料已经采用jieba分词并保存成tsv格式(你也可以采用其他工具分词,在torchtext里面分词速度较慢)
without any pre-trained word embedding
torchtext-tutorial
Convolutional Neural Networks for Sentence Classification