In Hungary the Central Office for Administrative and Electronic Public Services (KEKKH) is responsible for providing IT support for elections, referendums or nation-wide surveys and consultations. In 2012 for a nationwide consultation KEKKH decided to look for a sub-contractor who can provide a cost effective, automated and highly secure solution to replace the fully manual processing and data entry used before. Since the forms contained a lot of handwritten text, an exclusively automated, engine-based solution was incompetent due to the high error rate. The challenge was to find a secure, yet cost effective solution, that combines automated and manual processing and can scan and process over a 1 000 000 citizen interaction forms, containing typed and handwritten text, as well as check boxes, in approximately 2 weeks.
Main requirement toward the new solution:
- Scan and process over 1 000 000 forms in just 2 weeks
- Improve data confidentiality during the processing
- Speed up manual data entry
- Reduce operational and processing costs
- Filter out fake or badly filled out documents
OUR SOLUTION – INTELLIGENT FORMS PROCESSING AND DATA ENTRY
MPS designed and implemented a special forms processing solution that was able to combine automated optical character recognition (OCR) with manual processing of handwritten text for 100% accuracy. The new solution from MPS was fully scalable and able to securely process a few hundred thousand filled-out forms in a matter of days, utilizing the existing hardware resources.
- Envelopes are opened, forms are scanned
- Automated image fixing (rotation, cropping, etc.)
- Document filtering (fake, not properly filled-out, etc.)
- Text field recognition and automated processing (OCR)
- Manual processing of handwritten text (using crowdsourcing)
- Monitoring, real time statistics
After the envelopes were opened and the documents were scanned, the created images were pre-processed by the built in imaging software. Then, the automated recognition engine (OCR and OMR engines) processed the filled out forms, at the same time, the hand-filled parts of the forms were extracted and broken down to very small pieces (in many cases single letters or numbers) and sent out to data entry users, to type in though a special, secure web interface. This guaranteed the confidentiality of personal data, because the data entry users received only a few anonymous letters or digits meaningless on their own.
The results of the automated processing and the manual data entry, were put back together and exported into the necessary database. Parallel processing, built into the solution, guaranteed highly scalable and effective processing without the need of any complex infrastructure. Detailed history logs, real-time results and statistics were available during the whole process. This custom forms processing solution became the base for our present, award winning, document and forms processing solution – MPS IntelliVector.
- Automated processing of 1 000 000+ forms in 2 weeks
- Same-day processing of incoming forms (up to 100 000 pages/day)
- Accurate processing of handwritten text
- 100% confidentiality of personal data
- Automated filtering (fake, incorrect documents)
- Real-time statistics
- No extra infrastructure needed