Automated Test Case Generation from Input Specification in Natural Language

This paper studies the problem of automated test case generation for online coding test, i.e., given an input specification in natural language, how can we generate test cases automatically to examine the correctness of the code implemented by the testee? To tackle the problem, this paper proposes an approach that first extracts noun phrases from an input specification; then it removes irrelevant noun phrases and only retains the key phrases related to input construction; by reorganizing these key phrases, it can form an information tree and generate test cases accordingly. We have evaluated our approach with two datasets from LeetCode and ACM and achieved promising results.