Source Data Set. For our experiments, deep-web source data was col- lected via Google Base. Google Base is a central repository where merchants can upload their databases thereby publishing the databases over the web. Google Product Search works over the Google Base. Google Base provides API-based access to data, returning ranked results. Google Base’s Search API for shopping allows querying of data uploaded to Google Base. Each deep-web source in Google Base is associated with a source-identifier (SID). For selecting sources for our multi-topic deep-web environment, we probed Google Base with a set of 40 queries. These 40 queries con- tained a mix of camera model names, book, movie and music album titles. From the first 200 results of each query, we collected the SIDs and considered them as a source belonging to our multi-topic deep web environ- ment. We collected a total of 1440 deep web sources for our multi-topic environment: 276 camera, 556 book, 572 movie, and 281 music sources.
