Additional Experiments on Negative Rules

Abstract

We implemented the General and Enhanced algorithms as described in our technical report "Generic Entity Resolution with Negative Rules" and conducted extensive experiments. We ran our experiments on a comparison shopping dataset provided by Yahoo!. In this application, hundreds of thousands of records arrive on a regular basis from different online stores and must be resolved before they are used to answer customer queries. Because of the volume of data, we used blocking techniques to partition the data into independent clusters and then applied our algorithms on each cluster. In our experiments, we used a partition containing records with the sub-string "iPod" in their titles; we will call these iPod-related records from now on. The algorithms were implemented in Java, and our experiments were run on a 1.8GHz AMD Opteron processor with 20.4GB of memory. Though our server had multiple processors, we did not exploit parallelism.