This thesis introduces DescribeX, a powerful framework that is capable of describing arbitrarily complex XML summaries of web collections, providing support for more efficient evaluation of XPath workloads. DescribeX permits the declarative description of document structure using all axes and language constructs in XPath, and generalizes many of the XML indexing and summarization approaches in the literature. DescribeX supports the construction of heterogeneous summaries where different document elements sharing a common structure can be declaratively defined and refined by means of path regular expressions on axes, or axis path regular expression (AxPREs). DescribeX can significantly help in the understanding of both the structure of complex, heterogeneous XML collections and the behaviour of XPath queries evaluated on them. Experimental results demonstrate the scalability of DescribeX summary refinements and stabilizations (the key enablers for tailoring summaries) with multi-gigabyte web collections. A comparative study suggests that using a DescribeX summary created from a given workload can produce query evaluation times orders of magnitude better than using existing summaries. DescribeX's light-weight approach of combining summaries with a file-at-a-time XPath processor can be a very competitive alternative, in terms of performance, to conventional fully-fledged XML query engines that provide DB-like functionality such as security, transaction processing, and native storage.