BioWordVec - how to handle phrases

[griff4692]

The MayoSRS and UMNSRS_similarity datasets mostly contain phrases. Did you use mean pooling to get the results you reported or some other pooling mechanism for ngrams longer than 1?

Thanks