A natural language interface to answers on the Web can help us access information more ef- ficiently.  We start with an interesting source of information—infoboxes  in Wikipedia that summarize factoid knowledge—and develop a comprehensive  approach  to  answering  ques- tions  with  high  precision.    We  first  build  a system to access data in infoboxes in a struc- tured manner. We use our system to construct a crowdsourced dataset of over 15,000 high- quality,  diverse  questions.   With  these  ques- tions, we train a convolutional neural network model  that  outperforms  models  that  achieve top results in similar answer selection tasks.