They have tackled plagiarism; next on the list is contract cheating
by Max Smolaks 27 September 2019
Cyprus-based startup Unicheck has released Emma, an AI agent that can verify the authorship of academic texts, hoping to stamp out cheating in universities.
Automated plagiarism checks in academia are nothing new: this journalist remembers the dread of summitting his university assignments to an automated system a decade ago. However, traditional anti-plagiarism tools are unable to detect instances when the student simply paid someone else to do the work – and thanks to the Internet, essay and even dissertation writing has become a booming industry.
It is hiding in plain sight too – a quick search for “essay writing” on popular odd jobs website Fiverr returns more than 2,000 offers for this frowned upon, but not illegal, service.
By using machine learning, Emma can pick up on the student’s individual style, vocabulary and sentence structure – helping weed out academic cheaters with a lot of spare cash.
“Pressure to succeed often pushes students to cut corners and resort to contract cheating. Academic writing services cover anything from compiling a PhD thesis to a one-page essay,” said Serhii Tkachenko, CEO at Unicheck. “We have released Emma to encourage students to develop their own voice.”
Unicheck (formerly Unplag) has been developing cloud-based plagiarism detection tools since 2014, comparing submitted texts against web pages, open source repositories, and the user’s internal library.
Using natural language processing (NLP) and stylometry technologies to verify authorship is a new direction for the company. The AI-based assistant requires three texts with a word count ranging between 300-1,000 to study the individual’s writing voice. It can then decide whether any further submissions were written by the same person.
Emma is available as part of Unicheck’s Originality Report, the software suite used by more than 1,000 educational organizations worldwide. The company says that no manual work or programming of any kind is required to configure the new algorithm-based system.