Buch, Englisch, Band 334, 272 Seiten
Buch, Englisch, Band 334, 272 Seiten
Reihe: Dissertations in Artificial Intelligence
ISBN: 978-1-60750-719-2
Verlag: IOS Press
In view of this discrepancy, this work proposes a dedicated framework for modeling proof granularity. Within this framework, techniques are identified, developed and combined that serve to adapt computer-generated proofs to a step size suitable in a proof tutoring scenario. Norms for proof granularity are represented explicitly and can be inferred from classifications by human experts via machine learning techniques. The detection of suitable levels of granularity is applied to the automated assessment of proof exercises and in the generation of proof presentations. An empirical study investigates the automated learning of granularity judgments from experienced tutors.