Download 90K+ word vectors (.txt format with zip compression) for unseen words that are not covered by Word2vec (Google News), but defined in WordNet 3.0.
Please find the details of the approach in the paper:
M. T. Pilehvar and N. Collier Inducing Embeddings for Rare and Unseen Words by Leveraging Lexical Resources. EACL 2017, Valencia, Spain. [pdf] [bibtex]