I'd recommend python and the NLTK package, which has a corresponding book [0] that will bring you up to speed on the major points without skimping on the conceptual details along the way. And there's a lot of good work going on in python, like pyTorch & TensorFlow.
But anyone looking to do serious work that skips the type of content covered in NLTK Book in favor of jumping in to DNNs is doing themself & their ambitions an extreme disservice.
[0] http://www.nltk.org/book/