Token embedding

[Bessem11]

I read your paper "Styler: Learning Formatting Conventions to Repair Checkstyle Errors " and it was really interesting. My problem is that I didn't find which method did you used to embed the tokens (one-hot encoding, word2vec ... ?) in other words which method did you used to transform the Buggy abstract token sequence into a vector in order to feed it to LSTM.

[monperrus]

Thanks for your question @Bessem11

The tokens of the buggy abstract token sequence are embedded by the seq2seq network (OpenNMT with bidirectional encoder). We don't use any pre-training or pre-computed embeddings.