YouTube

Got a YouTube account?

New: enable viewer-created translations and captions on your YouTube channel!

English subtitles

← Unit 22 04 Problems and Solutions Questionmp4

Get Embed Code
2 Languages

Showing Revision 2 created 05/25/2016 by Udacity Robot.

  1. Now, what do we learn from this? It seems like there may be some serious problems with this grammatical approach to language.
  2. And I want you to tell me which of these you think are serious problems, true or false.
  3. So, one issue is, is it easy to write a grammar that will emit good parses accidentally? Tell me if that's true or false.
  4. Is it easy to write a grammar that will include spurious parses, or bad parses?
  5. Is that true or false? And trees are unobservable compared to words, which are observable. Is that a problem, true or false?
  6. And then let's look at some possible solutions. Now, one possibility would be,
  7. if we have too many parses to deal with, and we don't know which ones should be the right ones,
  8. maybe we should associate a probability with each tree.
  9. Is that a good idea? True or false?
  10. Secondly, if words are put together that don't seem to make sense, maybe we should bring back word association statistics,
  11. like we had in the Markov models, and incorporate them into the tree models? Is that a good idea, true or false?
  12. And finally, if we have the problem of ambiguity of our trees, maybe we should make the grammar unambiguous.
  13. After all, that's what we do with our programming languages. We make our programming languages unambiguous,
  14. so that, for example, A plus B times C, we know, is parsed as if it had parentheses around the B and C,
  15. and is not ambiguous as to whether it could be parsed the other way.
  16. Would it be a good idea to make our grammar unambiguous?