Reason for scaling similarity scores

[jabalazs]

Hi! Thank you for creating this repo.

Could you please tell us why you decided to multiply the MTurk similarity scores by 2:

https://github.com/kudkudak/word-embeddings-benchmarks/blob/8fd04891a92d313cc3b6956a43f25c9e44022e0e/web/datasets/similarity.py#L37

and those of the RG65 dataset by 10/4?

https://github.com/kudkudak/word-embeddings-benchmarks/blob/8fd04891a92d313cc3b6956a43f25c9e44022e0e/web/datasets/similarity.py#L172

I skimmed both papers and couldn't find any references to this.

Thank you!