The main characteristics of our datasets and data synopses size are summarized in Table 1.
To demonstrate the efficiency of our Data Synopses Indexing( DSI) scheme, we implemented the traditional Inverted List Indexing(ILI) scheme in Berkeley DB and compared it with our indexing scheme.
1) Varying bestK value: We first fixed the height and width of data synopses and varied the bestK value, i.e., the number of documents returned, to measure the average precision and recall of a query.
2) Impact of Height Factor: In the second set of experiments, we fixed the bestK value and the width of data synopses to see the impact of different height factors on precision and recall.
3) Impact of Width Factor: Finally, we fixed the size of the bestK value and the height of data synopses to see the impact of different width factors on precision and recall.
In , the authors further propose a novel class of XML synopses, termed XCLUSTERs, that provides a unified summarization framework to address the key problem of XML summarization in the context of heterogeneous value content.