with chatGPT and all the other future systems, a new problem in the evaluation of (not only) student’ work has emerged: the verification of references. chatGPT is known for fabricating references that look real but do not exist. a system that goes through a text, extracts all the references (usually collected in footnotes or in a list of references at the end, but sometimes inline) and verifies that these references actually exist seems like an easy task. .
when looking at the problem from a computational point of view, verifying an reference (especially in the light of different formats and standards) is not a trivial thing. even parsing the title and author(s) from such a reference can be a hard problem to solve, but algorithmically asserting that it is an actual reference is quite a challenge.
if well done, such a system can also be used to in verify web pages that all the linked material still exists and points to the right pages.
with all these requriements, creating such a system sounds like a really good challenge for a bachelor project + »bachelorarbeit«.