|
Sort Order:
[
Permalink
| « Hide
]
Mary Gardiner added a comment - 13/Jun/09 02:48 PM
It would be even cooler to provide two or three word phrases of interest. Instead of doing this by simple frequency counts, for combinations of words you can use a statistic like pointwise mutual information http://en.wikipedia.org/wiki/Pointwise_mutual_information to try and find words that are occuring *together* more than their individual frequencies suggest.
I did a rough implementation of this during the Melbourne hackfest. In the coming week I shall clean up the code and create a project on github for it.
Awesome Stephen, I've assigned the ticket to you and also made you a 'developer' in the ticketing system so you can update tickets.
I'm looking forward to seeing the code! Cheers, Henare Took a bit longer than hoped, had to cut back on functionality but it is on github
http://github.com/srbartlett/words-in-parliament Look froward to making other contributions to this and other features. screen show showing trigram words for a given day.
Labs is almost up and running and so we'd love to put this in Labs.
I've tried to deploy this to Heroku but it's got some problems:
http://hollow-stone-20.heroku.com/ Yet to debug it properly. index.html contains references to localhost:4567
you might want to give that a tweak. That's it Stephen, thanks! (How did I miss that??)
Fixed in this commit: http://github.com/henare/words-in-parliament/commit/97921cc9b6f65ea04db683fe626972271dba3164 Will move on to putting it on Labs now. Will see what Tim comes up with to solve
Done. The experiment is now deployed to http://labs.openaustralia.org
If we want to move this feature ahead, I suggest we create another ticket. Also, if we want to create a ticket for the Capitol Words idea for another hackfest, we can create a ticket under OAF. |
||||||||||||||||||||||||||||||||||||||||||||
New Feature
Closed
Minor
Ideas for Hackfest: Australian version of http://www.capitolwords.org/