TF_LSTM_Text_Classify

Tensorflow+lstm+text_classify(support chinese text classification and variable batch_size)

#Network:

Embedding + lstm + mean_pooling + Variable batch_size

#Requirements

Python 3.5 (> 3.0)

Tensorflow 1.2

#Introduction

1. This is a multi-class text classification (sentence classification) problem.
2. This model was built with LSTM(lstm/bi-lstm) and Word Embeddings(word2vec) on Tensorflow.
3. It supports the variable batch size.(the batch size of test code(prediction) is 1)

(在训练和测试时,每个epoch样本被分成很多batches,最后一个batch的size小于batch_size时也是可以去训练和测试的,不用舍弃这些样本)

4. It supports Chinese text classification, but you need the pretrained word2vector model.

(通过word2vector训练中文的词向量)

5. I don't publish the data_helper.py , because you can write it according to yourself dataset.

(根据自己的数据集来写data_helper.py, 将数据集写到trainset 和devset两个变量即可,trainset 和devset中包括所有样本的数据和对应的label)

#python file

Mylstm.py: define the network: Embedding + lstm + mean_pooling

My_bi-lstm.py: define the network: Embedding + bi-lstm + mean_pooling

train.py: train the network

注: 欢迎指正讨论.