Document content analysis for digital archives

The critical ingredient to an effective digital archive is metadata. There are at least two major problems with metadata.  First, it has to be gathered from the data items, which can be expensive and error-prone. Second, the metadata always depends on some model for what should be recorded about items, but such a model is never complete enough to satisfy the scope of uses for the archive.  For these reasons, the key to releasing the potential of digital archives is automatic content analysis.

This talk will touch on the state of the art of automatic content analysis for scanned and electronic documents.  Both academic research and commercial applications are driving technology developments in this field.  At the current stage, however work on digital archives is not heavily resourced, so most of us with projects in this area fall under the category of "hobbiest".  We can still dream up scenarios and designs for systems that will enable our archiving projects, and we can scour the landscape for camera-based document scanners, OCR, doctype classification, and other technical elements we can assemble in our garages.