search

seraphinatarrant / embedding_bias

Repo for project on the geometry of Word Embeddings and how it influences bias downstream
4 stars 2 forks source link

Select Embedding Algorithms - FastText #1

Closed seraphinatarrant closed 4 years ago

seraphinatarrant commented 4 years ago

1st choice - FastText 2nd choice - GloVe

Proposed choices: word2vec, glove, fasttext, transformer encoder, etc

Everyone needs to agree on one to start with (a second one as a backup if time permits is fine).

pandyamugdha commented 4 years ago

FastText - will work better for Russian and Spanish than other algorithms. Using subwords is not a problem for English.

GloVe - in case we don't use other languages it makes sense to use whole words then.