Automatic Function Testing Algorithm Based On Deep Learning

Tester design and write test scripts based on software requirements described in natural language,but they need to be trained through structured programming or automated testing tool programming in advance.It makes labor cost high,and it takes more time to write scripts than the test case texts described by natural language.More time,test e ciency is reduced during the scripting phase.Test scripts lose the flexibility of natural language descriptions.Writing test scripts put more pressure on testers' professional expressions.Software requirement analysis based on natural language processing can e?ectively improve the automation of automated testing.Most of the existing Natural Language Processing did not perform well in solving Chinese natural language texts.Some of the results are only available in the form of intermediate software products or further improvement is needed due to the need for more interaction.In order to improve the above problems,this paper proposes an automatic function testing algorithm based on deep learning.First,the keyword sequence is extracted from software requirement specification of the Chinese natural language description by natural language processing based on the deep learning,The keyword sequence has special meaning for the operation steps.We define it as a triple that can highly abstract test operations,including actions,target page elements,and data parameters.Second we need to locate the target page element,crawl the page of the test system,get the HTML DOM tree of the entire page,and locate the page elements of the test operation according to the target page elements in the triple.Then we build an automatic testing framework based on Selenium+Test NG,and write keyword interface according to keyword sequence,which can realize the interface call of keyword sequence,automatically generate function testing script,drive the program to carry out automatic function testing,and finally generate the report of test results.In this way,the test engineer can locate the defects of the system under test through the test report,and does not need write a large number of test scripts manually,which can save the cost of sorting out the test results and improve the test e ciency.In our experiment,the accuracy of extracting keyword sequence from test requirement specification text described in natural language is 89.7%,which is about 7% higher than that of ATA based on segmented backtracking method.The automatic test case generation technology based on depth learning proposed in this paper is superior to the automatic test method of recording-replay,and less manual interaction is required.The success rate of converting natural language operation descriptions into automatic test execution is 4% higher than that of ATA;in addition,the execution time required for the technology is within acceptable limits.