Abstract
Crowdsourced testing is an emerging testing method in the field of software testing and industrial practice. Crowdsourced testing can provide a more realistic user experience. But crowdsourced workers are independent of each other, they may submit test reports for the same issue, resulting in highly redundant test reports submitted. In addition, crowdsourced test reports with multi-source heterogeneous information tend to have short text descriptions, but the screenshots are rich, and using only text information can lead to information bias in test reports. In view of this, this paper attempts to use the screenshot information in the crowdsourced test report to assist the text information to cluster the crowdsourced test report. Firstly, the text similarity and screenshot similarity of crowdsourced test reports are calculated respectively, then the similarity between crowdsourced test reports is weighted. Finally, test reports are grouped by clustering algorithm based on similarity measure. Testers only need to audit the test report as the representative, which greatly reduces the pressure of the tester’s report audit. The final experimental results show that the effective use of the screenshot information in the test report can achieve higher clustering accuracy.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have