Scoring Mechanism of Defect Report Based on Text Similarity

With the rapid development of the software industry in recent years, the demand for software testing talents in the society is also growing. Therefore, more and more colleges and universities set up "software testing technology" courses in related majors. Finding defects is a basic skill of software testing technology training. Some defects are difficult to find and some defects are easy to find. How to give reasonable evaluation according to the difficulty and quantity of defects in the defect report submitted by students often requires much of teacher's time and effort. This paper proposes a defect repetitive detection method based on the combination of TF-IDF and cosine similarity and proposes a defect report scoring mechanism based on the difficulty level and number of defects found by students. Comparing the manually scores of the defect reports by the teachers, the scoring mechanism is reasonable and effective, and greatly improves the work efficiency of the teachers.