How to parse postal addresses from documents

I have seen that google and yahoo do a neat job on parsing postal address information from free form text. What technique/api do they use? What can I use to parse all postal addresses from a pdf, word, text, html document.