Adjust the near-duplicate detection algorithm

[Ali Hürriyetoglu]

Take the size of the text into account while you calculate the similarity.

If they are below a certain token size, use another measure, for instance Gestalt approach.