Abstract:

We propose a new method to rank a special category of time-sensitive queries that are year qualified. The method adjusts the retrieval scores of a base ranking function according to time-stamps of web documents so that the freshest documents are ranked higher. Our method, which is based on feedback control theory, uses ranking errors to adjust the search engine behavior. For this purpose, we use a simple but effective method to extract year qualified queries by mining query logs and a time-stamp recognition method that considers titles and urls of web documents. Our method was tested on a commercial search engine. The experiments show that our approach can significantly improve relevance ranking for year qualified queries even if all the existing methods for comparison failed.