MB-01/Brainstorms
From ThorxWiki
(Difference between revisions)
(greatly expanded) |
m (3 revision(s)) |
||
(2 intermediate revisions by 2 users not shown) | |||
Line 1: | Line 1: | ||
− | === ElectricMonk had some ideas about [http://rhyme.lycos.com/ rhyming slang] === |
+ | === [[ElectricMonk]] had some ideas about [http://rhyme.lycos.com/ rhyming slang] === |
* Teach moby a whole bunch of lyrics |
* Teach moby a whole bunch of lyrics |
||
Line 13: | Line 13: | ||
=== [[Screwtape]] had some strange obsession about making it faster === |
=== [[Screwtape]] had some strange obsession about making it faster === |
||
− | ...like anyone's ever wanted a faster computer program. pfaw! |
+ | ...like anyone's ever wanted a faster computer program. pfaw! Using the [http://www.python.org/doc/current/lib/profile.html Python Profiler] would likely help. |
+ | |||
+ | Also: |
||
+ | |||
+ | * When somebody says something, have a 1 in 100 chance that Moby will pick a keyword and "markov for <keyword>" without any intervention. |
Revision as of 02:04, 25 October 2007
ElectricMonk had some ideas about rhyming slang
- Teach moby a whole bunch of lyrics
- Teach it to recognise some basic rhyming structures
- Have it use those as best as possible...
Nemo had some ideas about thesaurus
- Feed moby a thesaurus so it knows big lists of synonyms
Assume moby would be generating a sentance "A B C D E", It starts with word "A" and then sees "b` b`` b```..." as possible followon words from A. However, it checks the thesaurus and finds that Y is a synonym for A, and so looks for "z` z`` z```..." as valid words that have followed" Y. Thus any of "b` b`` b``` z` z`` z```" can legally be the nextword after A. This is a backwards synonym check, and probably relatively cheap.
You could also, but probably with more resources expense, do synonym checks on each of b` b`` b```, etc as well, and thus do forward synonym checks.
Screwtape had some strange obsession about making it faster
...like anyone's ever wanted a faster computer program. pfaw! Using the Python Profiler would likely help.
Also:
- When somebody says something, have a 1 in 100 chance that Moby will pick a keyword and "markov for <keyword>" without any intervention.