Event model allows memory-efficient processing of larger documents than will fit in memory - but has a memory leak evident when processing multi-gig XML documents.
No longer actively maintained - last update 2005. Not a huge problem, as it's feature-complete and mostly non-buggy.
For Java, JDOM and XOM are other DOM-like libraries. XOM aims to enforce correctness better than DOM4J, but is less easy to use. JDOM is very widely used, but older than and inferior to either XOM or DOM4J.
There are a number of Java StAX based pull parsers, of which Woodstox is currently the best. These are slightly less easy to use than DOM-style parsers, but more efficient.