Array top slicing in study.py  #11

On line 198 and 205 of study.py there are calls to extract_concepts_with_negation which are sliced for 1000 or 500 members.

My reading of the code is that this will arbitrarily limit the concepts retrieved from the documents to the first 500 or first 1000, which is ok for small data sets, but when I change it to larger values produces radically different results when using large data sets.

Can someone check this?