| With the development of China’s education career, all kinds of colleges and universities teaching management is becoming more and more onerous task, which makes professional teachers not having adequate time to engage in scientific research. The assignments and exams is an important part of teaching and learning activity. It’s inefficient that we continue to follow the way of manually correcting homework and generating test paper and it seriously affect the development of China’s scientific research career. In order to help professional teachers to be free from the onerouc teaching task and put more time to do scientific researches, we develop the Talent teaching assistant system. But herein elaborated is a subsystem called smart exam system.This thesis analyzes the current application and research situation of related systems at home and abroad in recent years. Combining with the school actual situation, we analyze the system’s main functions, the system’s performance and the robustness and scalability of the system. Then we complete the system’s design and implementation work. We have initially completed smart exam system’s development and testing work. This thesis also does the innovative research on the two key algorithms involved in the system, one of which is the text similarity algorithm used in uploading question and searching similar questions function, and the other is automatic test paper generation algorithm used in automatic test paper generation function. The text similarity algorithm takes into account both the efficiency and accuracy. On the side of accuracy, we give full consideration to the influence of professional keywords, commonly used keywords and keywords’property and give these key words different weight in calculating text similarity process. On the side of efficiency, we use text clustering by using inverted index to shrink computing similar text set. And we also use the way of filtering matrix-dimension to reduce the computational complexity to further improve the efficiency of the algorithm. For automatic test paper generation algorithm, we carefully analyze several commonly used automatic test paper generation algorithm such as the random sample algorithm, backtracking algorithm and genetic algorithm, but they don’t meet system’s request on real-time, efficient and simple aspects. Later, based on the CPU scheduling LT-backfilling algorithm, we design and implemented the LT-backfilling intelligent test paper generation algorithm. |
