Using Microtasks to Enhance the Recognition of Receipt Entities to Keep Track Of Expenses

Maximilian Altmeyer; Pascal Lessel

CHI 2016. Workshop: Productivity Decomposed: Getting Big Things Done with Little Microtasks, 5/2016.


Optical character recognition (OCR) has improved significantly in recent years and is used in various domains. However, OCR results are still error-prone and heavily depend on the quality of both the picture taken and the printed text. This makes it hard to achieve reliable OCR results in domains that are confronted with e.g. text fonts that are hard to recognize, pale ink, or crumpled paper. We investigated crowdsourcing and gamification to enhance OCR in the domain of receipt capturing to keep track of expenses and developed a budgeting app that allows for tracking expenses by taking pictures of receipts to extract relevant entities. To enhance the recognition algorithm, we used the outcome of different microtasks that were solved by users of our app.

Weitere Links

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence