This paper describes a text mining program that computes the Barthel score of functional status by analyzing clinical notes stored in Electronic Health Record systems(EHR) and comparing them to textual evidence provided by clinical experts. The program demonstrates high accuracy and overall reliability based on a relatively small number of expert-abstracted charts. It offers an efficient and affordable method for estimating functional status using clinical notes. An important feature is an architecture that facilitates interaction between users and the program, allowing the program to improve its performance based on user feedback .