IMaT: Unsupervised Text Attribute Transfer via Iterative Matching and Translation
- Enriched test set for Yelp: Test Set (200 sentences with 5 human references)
- Outputs of 5 models (ours, CrossAlignment, Multi-Decoder, Style-Embedding, Delete-And-Retrieve) on Yelp and Formaliy.
Paper
Our paper is accepted by EMNLP 2019 (long paper):
IMaT: Unsupervised Text Attribute Transfer via Iterative Matching and Translation. Zhijing Jin∗, Di Jin∗, Jonas Mueller, Nicholas Matthews, and Enrico Santus.