Support IR query refinement by partial keyword set

As our early work, we proposed a data mining model for query refinement using Association Rules (ARs) mined among keywords of a large document database. In this paper, first, we explicitly identify principles of a data mining model for query refinement, and show that our model satisfies the principles. Second, we propose a mechanism to reduce the large number of ARs using a concept called "prime keyword". The effectiveness of our approach will be discussed. Considering the fact that most of today\´s searching engines on the Internet are keyword based, our approach can be easily extended to support Web applications.