I use a variation of the method of Tom Leinster : I take a large size (A4) notebook with blank pages (no lines so that I can write big and draw pictures more easily) and take notes, but *no more than one page per hour*. It enables you to write down any non-standard definition, the most important references, and the main results (or a summary of them if they are too intricate) but forces you to extract the core of the talk.