And I quote: "Microsoft Research in partnership with Bing is happy to launch the Speller Challenge." First place is $10,000, starts Jan 17, 2011, contest is on May 27, 2011.
We have the baseline implementation of the required system in our tutorials, see Spelling Tutorial. The system is not wrapped in a web service and you will need to dig up your own training data—Wikipedia anyone?
There may be issues if you want to use LingPipe tools to build the system, here is our Royalty Free License. For the purposes of this contest we will create a free license for contest use that is non-restrictive. If you’d like to request a free license for the contest, send me an email at breck@lingpipe.com.
Breck
December 22, 2010 at 4:27 am |
The challenge is cute, but totally inoperational for improving the suggestions of bing, primarily due to inavailability of accurate training data.
As google have shown, the suggestions come from analysing masses of other queries and large scale query-webpage mappings. I seriously doubt that even the bes string-similarity-or-whatever algorithm is usefull in this kind of task.
December 22, 2010 at 6:45 pm |
You’re right that you won’t be able to build a competitive spell checker using only the data they’ll provide. More data means more accuracy. And it usually trumps everything else once you have a reasonably good model. In this, the Microsoft/Bing spelling challenge is just like the Netflix challenge.
One of the widely cited paper on the more data, more better lines is from Banko and Brill (of Microsoft Research at the time of writing):
On the other had, any improvement to string matching should help with spelling correction. Modern spelling correctors combine at least a model of what’s expected and how people are likely to make errors typing and spelling.
Here’s an explanation from Peter Norvig, director of research at Google:
http://norvig.com/spell-correct.html
I really like the following paper by Toutanova and Moore (also of Microsoft(at the time of writing) on using pronunciation to model brainos in spelling (along with other factors):
December 27, 2010 at 9:40 pm |
I would like to suggest you the writing tool that you may want to try… http://www.spellcheckertool.com/ where it’s not just a spell checker, but also a grammar checker, style checker, etc.
^_^
Source(s): http://www.englishwriting.net/features/
February 18, 2011 at 8:38 am |
MSR Web n-gram spell check TREC 2008 data down : http://tinyurl.com/6xry3or ? Unable to download !