Why Stanford CoreNLP over Scikit-Learn or NLTK?

Scikit won’t be a choice if your requirement is Natural Language Processing. It’s a good tool for machine learning, not NLP. Stanford core NLP is Java and NLTK is Python.

Are you making a decision based on this?

The real difference is when your requirement is high level. By high level, I mean syntactic or semantic level processing. Stanford core NLP is best suited for the semantic level. It’s NER classifier, dependency parsers and chart/functional parsers are great tools to work with grammatical functions(subj, obj, pred). NLTK doesn’t score well here. CFG parsers and the huge corpus make NLTK attractive for regular syntactic level of processing and text mining. Since Python has great libraries for ML, data mining, NLTK is an attractive choice for your text mining projects. If your requirement is core NLP then go for SNLP.