This paper explores two classes of model adaptation methods for Web search ranking: Model Interpola-tion and error-driven learning approaches based on a boosting algorithm. The results show that model interpolation, though simple, achieves the best re-sults on all the open test sets where the test data is very different from the training data. The tree-based boosting algorithm achieves the best performance on most of the closed test sets where the test data and the training data are similar, but its perfor-mance drops significantly on the open test sets due to the instability of trees. Several methods are ex-plored to improve the robustness of the algorithm, with limited success.

PDF file

In: EMNLP

Publisher: Association for Computational Linguistics
All copyrights reserved by ACL 2007

Details