Skip to content

Files

Latest commit

ecd8c88 · Oct 6, 2021

History

History
25 lines (18 loc) · 891 Bytes

File metadata and controls

25 lines (18 loc) · 891 Bytes

Applying the Transformer to Character-level Transduction

Shijie Wu, Ryan Cotterell, and Mans Hulden. Applying the Transformer to Character-level Transduction. EACL. 2021.

Experiments

We use morphological inflection as an example since the data is hosted on GitHub. Feel free to contact shijie.wu_at_jhu.edu for g2p and transliteration. The historical text normalization dataset can be downloaded here

# We use latin as an example
lang=latin

# download this repo
git clone https://github.com/shijie-wu/neural-transducer.git
cd neural-transducer

# download data
mkdir data
cd data
git clone https://github.com/sigmorphon/conll2017.git
cd ..

# Run feature-invariant transformer (`arch=tagtransformer`). For regular transformer, `arch=transformer`.
sh example/transformer/trm-sig17.sh $lang