Maarten Balliauw has written about a method he's developed to convince the Zend_Search_Lucene component of the Zend Framework to index the contents of a Word 2007 document.

Lucene basically is an indexing and search technology, providing an easy-to-use API to create any type of application that has to do with indexing and searching. If you provide the right methods to extract data from any type of document, Lucene can index it. [...] Sounds like a challenge!

He works through the three step process to getting the searching working, the key being his readDocXContents() function that goes through the Word file and returns all the text it can find. This is passed back out so the Zend Framework component can pull the data in and search (their example is on the string "Code Access Security").

You can grab the the full code here.