sara's use case

reduce to 80 novels
datastore = Norstad
use WordHoard to define training set
SEASR and WEKA tools to analyse the work set
list of stop words
sorted list of predictions nora without and interface

dat move to SEASR
output: list of feature with weightings

repoint Stefan's workbench to new data structure
Is the proxy outside or inside the loop
proxy to remain for now
relationship of SEASR to datastore

proxy piece that speaks to John's data store
ditch nora and start with new calls
modify calls to respond to greater richness of Monk datastore

How does Stefan's workbench

API will change with data model