A quantitative approach to the semantic classification of English blends with token-based semantic vector space modeling