I will summarize the current status of the http://keywords4bytecodes.org project which seeks to predict relevant English terms from compiled Java code. By leveraging large collections of readily available open source programs, I have built a comparable corpus of 1M pairs of dissassembled Java methods and their corresponding Javadoc comments. In this talk I will present some preliminary results using an IBM Model I and discuss its application to Reverse Engineering, as presented at the Reverse Engineering convention (http://recon.cx) last summer in Montreal.