I saw a tweet recently that pointed to: Learning to classify text using support vector machines, which is Thorsten Joachims’ dissertation, The Maximum-Margin Approach to Learning Text Classifiers as published by Kluwer (not Springer).
Of possible greater interest would be Joachims more recent work found at his homepage, which includes software from his dissertation as well as more recent projects.
I am sure his dissertation will repay close study but at > $150 U.S., I am going to have to wait for an library ILL to find its way to me.