is it feasible to create loose templates for where the data is and extract that way? i have a mothballed project that did pretty well. it was able to discern different templates from a mass of documents.