Open juditacs opened 9 years ago
The currently used embedding (word2vec default) assigns high similarity to non-similar words. We should try to find an embedding that is stricter with unsimilar words.
The currently used embedding (word2vec default) assigns high similarity to non-similar words. We should try to find an embedding that is stricter with unsimilar words.