Why not use a full-text indexing engine?
Full text indexing engines are very fast and appropriate for performing broad queries across a set of documents. The drawbacks to using them for XML applications is that creating an inverted index causes you to lose document context, requiring very intensive post-processing to isolate the document nodes that are being managed. Xindice provides efficient indexing of element and attribute values in addition to the ability to return and manage aggregated document fragments.
