The proposal aims to develop a framework to fuse the outputs of multiple open-source OCR packages (gocr, tesseract, ocrad, ocropus) to improve recognition accuracy over individual packages. The framework will investigate techniques like boosting, cascading, and adaptive fusion. In addition to machine-generated text, the project will collect a dataset of handwritten text samples gathered electronically using a tablet PC to test performance on a wider range of data. Formal comparisons will evaluate character and passage recognition. If successful, the framework could provide a low-cost, high-performance alternative to commercial OCR systems.