Korean to English Translation

Translation task implementation with transformer and seq2seq attention mechanism with low level pytorch code

Tokennizer

• Korean : pip install konlpy
• English : pip install nltk

Architecture

BiLSTM Seq2Seq with Attention

• Word Embedding
• BiLSTM Encoder or Stacked LSTM Encoder
• LSTM Decoder with Attention mechanism

Transformer

• Positional Encoding + WordEmbedding
• Multi-head Attention, Position-wise Fead Forward Network
• ResNet + NormLayer

Refer

• Attention mechanism : Neural Machine Translation by Jointly Learning to Align and Translate (https://arxiv.org/abs/1409.0473)
• Transfomer: Attention is all you need (https://arxiv.org/abs/1706.03762)

Start

pip3 install -r requirements.txt