Title :
Anomalies detection in evaluation of essay type answers by multiple evaluators
Author :
Shukla, A. ; Chaudhary, B.D.
Author_Institution :
Dept. of Comput. Sci. & Eng., Motilal Nehru Nat. Inst. of Technol. Allahabad, Allahabad, India
Abstract :
In this paper, we present a methodology to detect suspected anomalies in evaluation of essay type answer books based on relationship between marks awarded and symbols used during evaluation. We identified dominant symbols used by evaluators and their relationships with marks awarded based on analysis of evaluated essay type answer copies. Further, for each evaluator relationship between symbols used and marks awarded is distinct. We identified outliers for each evaluators and each subject based on relationship between marks awarded and tick symbol and termed them as suspected anomaly. All grievances of students related to evaluation were in the set of identified outliers.
Keywords :
educational administrative data processing; security of data; anomaly detection; dominant symbol; essay type answer book; multiple evaluators; Computer science; Conferences; Correlation; Correlation coefficient; Educational institutions; Linear regression; anomaly; assessment methodology; essay type answer; evaluation; grievances; outlier;
Conference_Titel :
Teaching, Assessment and Learning for Engineering (TALE), 2013 IEEE International Conference on
Conference_Location :
Bali
DOI :
10.1109/TALE.2013.6654413
