Green ICR: Semi-Automated Census Record Indexing with Emphasis on Human Computer Interaction

Human-based computation is an approach that utilizes the abilities and strengths of both humans and computers to achieve a symbiotic interaction that is stronger than either agent in isolation. We propose a system that amplifies the capacity of a human indexer by adding an intelligent handwriting recognition engine to the indexing process. This recognition engine will learn patterns in handwriting as the indexer works, and then will amplify the indexer by automatically labeling handwriting with similar patterns. The recognition engine may also prompt the user to label examples that will best help it to learn. Preliminary results show that applying handwriting recognition technology could significantly reduce the number of fields an indexer is required to hand label. As a byproduct, we also believe the proposed system will be more interactive and more enjoyable for the indexer.