Most complex analyzer which can identify special tokens (e.g., proper names, email addresses, etc), normalize text, remove stop words (i.e., frequently occurring, but unimportant words like "a", "an", "the", etc) and punctuation.
Use Stanford CoreNLP (http://nlp.stanford.edu/software/corenlp.shtml) for part-of-speech tagging. CoreNLP also has information retrieval functionality like named-entity-recognition and can do sentiment analysis