Computational Music Analysis requires data in computational form to work with, such as MusicXML or MIDI. Currently this is often derived by hand. Software to read information from scans of scores does exist but it typically operates at a level of accuracy much lower than is now the case for analogous OCR software reading text. This paper will report results from a project to improve the accuracy of OMR by making use of multiple sources of information. Scans of different editions of a work, and sometimes multiple scans of the same edition or scans of scores and parts, are now available from sources such as IMSLP. Our hope is that the increase in accuracy which will result from the combination of information from multiple sources will make OMR a practical tool for computational music analysis.