Paper - iPres 2013 - Lisbon

Users frequently mistype queries and blame the web archive for poor search results. The addition of a query suggestion functionality in the Portuguese Web Archive had great impact on the perceived quality of the service. In this work, we tested five existing solutions over two datasets. However, existing solutions do not work well, because they rely in predefined lexicons to detect misspellings. We improved the best solutions with a set of rules automatically tuned with an index of archived web collections. The final result can be tested at http://archive.pt and the software is publicly available as an open source project.

application/pdf