An Empirical Evaluation of Simple DTD-Conscious Compression Techniques James Cheney   View Paper (PDF)   Return to Paper Session 3: XML Abstract In previous work, we developed a streaming XML-conscious compressor xppm (or XMLPPM), and showed that it provides compression superior to other text compression and XML-conscious compression techniques then extant. The purpose of this paper is to investigate whether DTD information can be used to improve compression in xppm enough to justify the added implementation effort. In this paper we describe dxppm, a version of xppm that simultaneously validates and compresses XML relative to a DTD. We assume both sender and receiver have access to identical copies of the DTD. Because xppm already compresses both XML structure and text very well, and because of the complexity of the underlying PPM algorithms, it is easy to generate ideas for DTD-based compression that work well ``on paper'' but are either incompatible with xppm or do not improve compression relative to xppm. Our contribution is the development of four simple DTD-based compression techniques that do work well with xppm, validated by experiments showing substantially improved compression for a variety of real data sources. ©2006 Association for Computing Machinery