In PairSimilarity.java, Method cookLexicalItem() with parameter textType being SURFACE or SURFACE_TAGGED, the for loop is executed for each single word in the text, although 1 time would be enough.
In long texts this results in a really long execution time.
Instead, at case SURFACE or SURFACE_TAGGED, it could just get out of the for loop on execution.
(By comparing 2 texts with ~400 words, this reduces the calculation time from 3min20secs to 35 secs, with the same output.)