For SVO, dependency parsers such as the Stanford Parser http://nlp.stanford.edu:8080/parser/ or MATE tools https://code.google.com/p/mate-tools/ are better than constituency parsers, since the subject and the object of the verb (as in SVO) are directly labeled as such. They should work for NL questions, however if you use this inside a QA system, additional processing might be advantageous to get the focus of the question, cf. http://researcher.watson.ibm.com/researcher/files/us-heq/W%284%29%20QUESTION%20ANALYSIS%2006177727.pdf