Non-verbal Korean Dialogue Modeling Using Deep Learning

Summary

Korean conversation modeling using deep learning. We build Seq2Seq Model and HRED Model, and test them.
We completed the chatbot by linking the results with Kakao Talk Chatbot API.

Paper Study

1. (Word Embedding 1) Distributed Representations of Words and Phrases and their Compositionality
2. (Word Embedding 2) GloVe: Global Vectors for Word Representation
3. (GRU) Learning phrase representations using RNN encoder-decoder for statistical machine translation
4. (seq2seq) Sequence to Sequence Learning with Neural Networks
5. (attention) Neural Machine Translation by Jointly Learning to Align and Translate
6. (conversation) Building End-To-End Dialogue Systems Using Generative Hierarchical Neural Network Models

Team member

• Daeun Kim
• HaaYeon Lee
• JiHun Suk
• EunJeong Jang
• JaeYeol Jeon

Requirements

• python 3.6.5
• tensorflow 1.9.0
• nltk 3.3

References

• http://gom.gomtv.com/main/index.html?ch=subtitles&pt=l&menu=subtitles (GomPlayer)
• https://ithub.korean.go.kr/user/total/database/corpusManager.do (National Korean Language Institute)