The effect of using different lexicon (synonym relation) sizes on the proposed method with SKB, NNE and HNE evaluated on SCWS dataset. The full ukWaC corpus is used as the corpus.