Tensorflow2.0-examples

Implementations of deep learning model with Tensorflow 2.0

Overview

This repository is specialized for neural response generation or dialogue generation in the field of NLP.

Models

This repository consists of various models related with dialogue generation:

1. Sequence-to-Sequence Model
2. Sequence-to-Sequence Model with Attention Mechanisms (Bahdanau & Luong)
3. Hierarchical Recurrent Encoder-Decoder (HRED) -> Not yet
4. Latent Variable Hierarchical Recurrent Encoder-Decoder (VHRED) -> Not yet
5. Transformer -> Not yet

To Do

• Applying the earlystopping technique
• Applying the beam search
• Tokenizing sentence at morepheme level
• Use other datasets (e.g. DailyDialog)

Datasets

To train and evaluate several models, we used korean daily chat data. You can be publicly available to download dataset by following this link.

Requirements

• Install tensorflow 2.0

$ pip install tensorflow

• Install ChatSpace

$ pip install chatspace

Reference

• Tensorflow 2.0 Neural machine translation with attention
• Tensorflow 2.0 Encoder-Decoder
• ChatSpace