It's a longstanding problem that a very large number of parse trees (1000s) can be found for sentences of any real complexity: there are so many of them that the one with the highest probability of being right is still likely to be wrong!