A new function in the admissions system can shorten the processing time for international applicants. A module based on artificial intelligence (AI) helps the system to automatically recognise passport copies.

”We hope that in the long term, this will lead to shorter processing times for applicants without a Swedish personal ID number,” said Robert Åhlén, system administrator at the Swedish Council for Higher Education (UHR).

UHR’s admissions system manages applications to courses and programmes submitted at Antagning.se and Universityadmissions.se. International applicants may be required to pay application and tuition fees in Sweden and must provide a copy of their passport when they apply for admission. This is particularly important for citizens of the EU, EEA and Switzerland who must document their citizenship in order to be exempt from paying fees.

Robert Åhlén, system administrator at UHRUntil now, admissions officers have had to manually check if the application contains a passport copy but thanks to a new module, this step can be done automatically.

”Before UHR can start processing an application, we have to check if the international applicant is required to pay application and tuition fees. We’re hoping this process can be shortened so that applicants can more quickly find out if they’re required to pay, and if they meet the entry requirements” said Robert Åhlén.

300,000 documents per year

The admissions system manages a large volume of documents each year – last year approximately 300,000. Document review takes a lot of time and was earmarked as a prioritised area for development. In the autumn of 2017, UHR began to work on how the process could be improved.

The IT Support and Systems Development Unit (ITS) at Umeå University develops and manages the admissions system for UHR. ITS contacted four students in the university’s computer science programme. As a part of their degree project, the students explored various AI solutions for the automatic recognition of documents. The first step was to list the types of documents the system manages.

”Passport copies in particular take a lot of time as the admissions officer has to review them manually. Passports also look a certain way regardless of which country they’re from. We decided it’s the type of document that a computer could learn to recognise” said Robert Åhlén.

Training with a large volume of passports

In order to learn to automatically recognise passports, the module needed large volumes of them to practice with. As a part of the students’ degree project, they fed the module with passport documents for several months, with positive results. As a summer job, the students were given the task of continuing to work with the development of a pilot module.

When the AI module was put to work with actual admissions documents, it could with an accuracy of 99.43 percent sort out passports. The module will be officially launched when master’s students begin submitting their applications in December and January.

Authenticity checked manually

Passport authenticity, expiration date and issuing country must still be checked by an admissions officer; the process is not yet completely automated. But if the new functionality works well, Robert Åhlén sees an opportunity to further develop the technology.

”The next step is to teach the computers to discern passports from different countries. It would be helpful to quickly sort out passports from EU and EEA countries and Switzerland, as those students aren’t required to pay fees. We need to determine their fee status as soon as possible so that their application can continue through the review process.