Identical scores with and without word2vec embeddings

[cosmin-novac]

Interestingly, the scores are nearly identical, whether I use a word2vec embedding model or not. Could someone explain how this is possible?

(excuse the german :) )