Algorithm recognition for programming tutoring

Abstract

Plan-recognition with template matching shows reasonable performance for recognizing general control structures. However, problems containing a well-defined algorithm such as sorting and searching are difficult to recognize by just the plan recognition methodology alone, because an algorithm is often highly optimized and therefore hard to divide into smaller meaningful units. Based on this observation, we propose an algorithm recognition methodology to augment the plan-recognition approach. Our algorithm recognition approach is based on the flow graph parsing which performs a partial recognition of functions. The methodology is extended to completely understanding algorithm implementation by providing the information of the program``s goal, which is available as a specification of programming assignment in our tutoring environment. Our effort for utilizing goal information is focused on two aspects; extending the role of transformation rules to represent algorithm-dependent information and looking for salient graph parts for an algorithm plan to determine which plan has been used to implement the goal. Transformation rules represent an alternative viewpoint of a subgraph. If an interpretation of a graph is suspended, alternative interpretation is tried on the transformed graph from different viewpoint. Identification of the algorithm plan in an implementation is performed by finding a discernible subgraph, clue, that uniquely determines which has been used among those plans for the algorithm goal. Preliminary evaluation was performed on 72 programs of C programming language containing sort algorithm, which were submitted by freshmen of C programming language course at KAIST. 55 program were recognized completely. Excluding 8 extraordinary programs, our methodology showed high recognition rate (55/64).