Met with MT to plan work on improving the ISE search. The biggest barrier as far as I can see is the sheer complexity of the scarcely-documented tagging system, combined with the range of different textual components that need to be separately searchable. We know we want to deploy a snowball analyzer, which means we'll probably have to replace the kwic module in exist with our own, but before we get that far, we need to test a range of search strategies to see how fast they can be. One approach I'm starting on is to generate very simple versions of all of the files designed for specific search targets (lines of plays/poetry; metadata; etc.). The process of creating these is helping me to get familiar with the ISE encoding, which will help later one whatever approach we end up taking. Ultimately, we should also be looking at producing an archive version of the core encoding for each play using TEI, and it seems to me that the current representation could perfectly well be converted to TEI, and would be a lot simpler as a result.