Microsoft Research and Bing are jointly hosting the Speller Challenge. The goal is to build the best service that could propose alternative spellings for search queries submitted to Bing. Entries must be submitted for the challenge in the form of a REST-based web service, and they will be judged based on their expected F1 score against a test set sampled from real Bing queries.

For development purposes, they are making available a TREC evaluation dataset through their Web-NGram service.  Refer to this page for detailed evaluation measures and REST service specs.

Time to start implementing!

Related posts:

1. CFP Semantic Web Challenge and Billion Triples tracks
2. ICWSM 2011 Data Challenge with 3TB of social media data
3. Microsoft rumored to buy semantic search startup Powerset
4. UMBC team 2nd in DARPA Shredder Challenge
5. The Semantic Web billion triples challenge at ISWC 2008