Research On Cognitive Diagnosis Of Students' Program Errors Based On Crowdsourced Feedback

The crowdsourcing model has brought more and more crowdsourced systems into people's perspectives,crowdsourced systems such as Amazon Mechanical Turk(AMT),providing a platform for a host of questions answered by Internet staff.It turns out that such a system is useful for solving problems that computers can not handle,such as sentiment analysis,image recognition and sentence translation.Following common crowdsourcing platforms such as AMT,more scholars in recent years have focused on crowdsourcing applications that address a specific issue.In a new programming language learning process,not difficult to find Most people are still in the programming process encountered a mistake even by relying on the standard editor,if we can improve this phenomenon,will be in the programming language Grasp the quality of a substantial increase.Therefore,this paper presents a CrowdCC(Crowdsourcing Code Correcting)system with practical significance for programming language beginners.Throughout the research process using Java programming language as an example,CrowdCC using web analytics to obtain a wide range of Java bug types as a large collection of java code snippet.This article designed and implemented three main modules for CrowdCC:Task Management,Task Release,Task Assignment.The first two modules support task requester free to add java code error correction,management history release tasks,the task posted to the xTurk site in an optional mode.In the Task Assignment module,the Reliability-A ware Task Assignment(RATA)algorithm is proposed.Compared with the other two task assignment algorithms,the algorithm proposed in this paper can make use of the feedback information of past staff on the history publishing task on CrowdCC.Give a better assignment scheme.