This set of homonyms is one I didn't think I had difficulty with. Then I realized it's simply because I don't use them that often in my work. A good review of the definitions over at Ginger Software made sure I'll keep them straight in the future.
Following a rather rambling train of thought after I played around with this one left me wondering: If the majority of closed captioning is done via software and not humans typing, what are the odds the computer will pick the correct spelling when the words sound the same. Is there an algorithm for context?