Evaluating OpenNLP vs NLTK

I am evaluating OpeNLP vs NLTK for Natural Language Processing. I have heard that NLTK is slower when processing, but not much else in the manner of comparison. Any specific reasons you may need to choose one over the other (in your experience) would be appreciated...
Using in combination with Lucene, SOLR, so I am not using this for ngramming, etc..
but will sit on top of this (or behind) for sentimental analysis etc... things harder to conceptualize just in indexing structures.

My environment is alrady poolyglot so python vs java is not really the consideration. I am just parsing as I am not real enthused about going full UIMA and supporting multiple frameworks (though a/b between them is enticing)

Both OpenNLP and NLTK are quite robust roll kits and can perform most NLP tasks. Tho I could prefer going with OpenNLP in your case because you are already using Lucene and Solr and integrating openNLP to a Java based stack is a lot easier then integrating NLTK to it, even tho you have a polyglot having to go through less pain I'd always better.

Performance is also decent concern when you are going to use it for search. NLTK anyways is quite old now and is probably not the best option today for such tasks, SpaCy is quite popular these days.

Also, maybe you could have a better luck asking this on stack overflow and not bhw