Yeah, Haha. 
Do you think a predictive intelligence model for similarity of entered text against completed/accepted records would be a good solution instead? It may be a bit opened-ended though. 🙂