Winning spelling algorithm
Written by Lucy Black   
Sunday, 24 July 2011

There were over 300 participants in the Microsoft Research/Bing Speller Challenge. The winners were selected automatically and were presented with prizes at a workshop last month.

 

spellerchallengebanner

 

The problem that this Microsoft challenge set out to tackle was getting a better set of results from search queries that could include mistyped or misspelled words - the subtlety, as explained in last December's blog post announcing the competition being that:

"One person's spelling error could be another's perfect query".

The contest was open to researchers and students worldwide and the blog post reporting the winners notes that there were entries from every continent except Antarctica.

For the purposes of the contest participants were given access to real-world data at web scale by using the Microsoft Research Web N-gram Services and were able to improve their algorithms and see how it compared to other spelling correction systems by using an evaluation service provided by Microsoft Research.

Given that the task was to devise an automatic algorithm, the winning entries were also selected automatically based on performance in figuring out the best spelling alternatives, for example, "Britney Spears" for "briteny spears".

Prizes were awarded as follows:

  • First place (US$10,000): Gord Lueck – Canada
  • Second place (US$8,000): Yanen Li, Huizhong Duan, and ChengXiang Zhai – United States
  • Third place (US$6,000): Yasser Ganjisaffar, Andrea Zilio, Sara Javanmardi, Inci Cetindil, Manik Sikka, Sandeep P. Katumalla, Narges Khatib, and Chen Li – United States
  • Fourth place (US$4,000): Dan Ştefănescu, Radu Ion, and Tiberiu Boroş – Romania
  • Fifth place (US$2,000): Yoh Okuno – Japan

In the video below Harry Shum, corporate vice president of Bing, explains the idea behind the contest and introduces the overall winner Gord Lueck who tells us how he set about the challenge.

 

speller


Related articles:

$30,000 for a better spell checker

Microsoft Web N-gram Services go public

 

Banner


Rust 1.82 Improves Apple Support
24/10/2024

Following Rust's six-week release cycle, version 1.82 has been released with higher level support for Apple, and a new Info subcommand for Cargo.



Mojo - Python Superset For Go-Faster Hardware
07/10/2024

Mojo, the new language from Chris Lattner and his AI-focused company Modular, has entered the top 50 of the TIOBE index, even though it it was only launched in 2023.


More News

Last Updated ( Sunday, 24 July 2011 )