We are releasing the misspellings dataset consisting of a set of pairs (wm, we), where we represents a correctly spelled word and wm is a misspelling of we. For more details about the dataset, please check our paper Misspelling Oblivious Word Embeddings published at NAACL 2019.
Bora Edizel: edizel@fb.com
Aleksandra Piktus: piktus@fb.com
moe is Attribution-NonCommercial 4.0 International licensed, as found in the LICENSE file.