FlashWB / finch

Deep NLP in TensorFlow r1.6

CPU Installation

$ pip3 install tensorflow==1.6.0 sklearn scipy

---

Contents

• Word Embedding（词向量）
• Text Classification（文本分类）
• Text Generation（文本生成）
• Text Control（文本控制）
• Sequence Labelling（序列标记）
• Sequence to Sequence（序列到序列）
• Memory Networks（记忆网络）

---

Word Embedding（词向量）

• Skip-Gram     Model     Test     Result

• CBOW     Model     Test     Result

Text Classification（文本分类）

• CNN     Model     IMDB Test     Result

  • CNN + k-Max Pooling     Model     IMDB Test     Result

• RNN     Model     IMDB Test     Result

  • RNN + Attention     Model     IMDB Test     Result

• Only-Attention

  • Simple Attention     Model     IMDB Test     Result

  • Scaled Dot-Product Attention     Model     IMDB Test     Result

• TFIDF + Logistic Regression     Model     IMDB Test     Result

Text Generation（文本生成）

• Char-RNN     Model   |   English Story Test   |   Chinese Address Test

  • Char-RNN + Beam-Search     Model   |   English Story Test   |   Chinese Address Test

• Character Aware CNN-RNN    Paper     Model     PTB Test

• Only-Attention     Model   |   English Story Test     Result   |   Chinese Address Test     Result

• CTC

Text Control（文本控制）

• VAE + Discriminator + Wake-Sleep

Sequence Labelling（序列标记）

• Bi-RNN + CRF     Model     POS Tagging Test     Chinese Segmentation Test

• Only-Attention + CRF     Model     POS Tagging Test     Chinese Segmentation Test

Sequence to Sequence（序列到序列）

• Seq2Seq     Model     Sorting Test

  • Seq2Seq + Attention     Model     Sorting Test

  • Seq2Seq + BiLSTM Encoder     Model     Sorting Test

  • Seq2Seq + Beam-Search     Model     Sorting Test

  • Seq2Seq + BiLSTM Encoder + Attention + Beam-Search     Model     Sorting Test

• Attention Is All You Need

• Pointer Network     Model     Sorting Test

• VAE

Memory Networks（记忆网络）

• End-To-End Memory Network

• Dynamic Memory Network

---

How To Use

1. The following command clones all the files (it is large due to historical reasons, sorry)

   下面命令会拿下所有文件（因为历史原因，所以体积比较大，不好意思）

   git clone https://github.com/zhedongzheng/finch.git
   

2. Use contents to find the model and test that may interest you, click on that test

   用目录找到感兴趣的模型和测试文件，点击它

   

3. Find the test file path

   发现测试文件的路径

   

4. run on command line

   在命令行运行它

   cd finch/nlp-models/tensorflow
   python rnn_attn_estimator_imdb_test.py
   

---

Questions Have Been Asked

• Is the function "add_encoder_layer" in "seq2seq_ultimate.py" correct?
• in vae code,how to use LSTMcell?

---