The advancement of technology has made learning possible regardless of temporal and spatial barriers through e-learning platforms. They host numerous MOOCs that deliver quality content to learners around the world. Most traditional materials used in classrooms such as lecture videos and readings can be scaled up to accommodate an unprecedentedly large size of students in a MOOC, but scaling up for evaluation and assessment becomes difficult, particularly when evaluation and assessment require human interpretation and judgement. To address this problem, instructors in MOOCs resort to peer assessment to tackle the enormous scale of learning evaluations. As peers have different understanding towards the concepts that appear in the assessment, how they are assigned to peer homework for asse...[ Read more ]

The advancement of technology has made learning possible regardless of temporal and spatial barriers through e-learning platforms. They host numerous MOOCs that deliver quality content to learners around the world. Most traditional materials used in classrooms such as lecture videos and readings can be scaled up to accommodate an unprecedentedly large size of students in a MOOC, but scaling up for evaluation and assessment becomes difficult, particularly when evaluation and assessment require human interpretation and judgement. To address this problem, instructors in MOOCs resort to peer assessment to tackle the enormous scale of learning evaluations. As peers have different understanding towards the concepts that appear in the assessment, how they are assigned to peer homework for assessment could affect the overall grading quality. In this thesis, we study an assignment problem in peer assessment. We aim at maximizing the learning concept coverage when homework is assigned to learners, which turns out to be an NP-hard problem. We propose a greedy algorithm and two variations for this problem. Finally, we conducted experiments on several datasets to verify the effectiveness of our algorithms.