Skip to content

paulx3/keras_seq2seq_word_level

Repository files navigation

Keras seq2seq word-level model implementation by wanzeyu

Overview

Keras implementation for seq2seq

In this project I try to implement seq2seq word level model using keras.

Resource Used:

  1. MSRP paraphrase corpus

Requirements:

  1. Keras
  2. Numpy

The reason I open this repo

  • The official tutorial about implementing seq2seq model is character level. But word level implementation is common.

QuickStart

python basic_seq2seq.py

Resource

  • test_source.txt original sentence
  • test_target.txt translation sentence
  • train_vocab.txt vocabulary

Problems

  1. The inference function has some problems

Progress

  • Write the evaluation code
  • Add Attention function
  • Using Bidirectional LSTM in encoder
  • Implementing Joint Copying and Restricted Generation for Paraphrase

References

About

Implementation of seq2seq word-level model using keras

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages