This is pretty much the main idea behind another paper that Eamonn wrote [0]. The idea is that you can normalize and discretize timeseries data into a lecicographical representation, then perform all sorts of interests analyses on it. For example, searching for common subsequences can be done via regular expressions.