The Verification Technique Of Complex Tasks' Non-functional Specification In Software Crowdsourcing

Crowdsourcing is a business model using the Internet to assign work,find creative ideas and solve technical problems.Crowdsourcing enables enterprises to achieve the goals at a low price on a global scale.There are many successful software crowdsourcing platforms.However,there are still many defects in the existing software crowdsourcing platforms.In general,requester often post simple decomposed subtasks rather than a whole complex task.Because it's easier to match workers.It always takes much to guarantee the consistency between the original and reconstructed tasks.Even though,it is often the case that requesters can't get the whole qualified software from all subtasks completed as required.To solve these problems,we present a solution to automatic verification of non-functional specifications' consistency between the original complex task and the output integrated by subtasks.First,an extensible and effective model is proposed to support the tasks' description.Considering the variety of non-functionalities,we provide not only common non-functionalities' meta-models,but also interfaces for requesters to define new non-functionalities' meta-models.Based on the meta-models,a description model for non-functionalities is provided.In the description model of tasks,different weights are used to represent different priorities of non-functionalities.On the basis of WF-net,the description model of the structure between tasks is proposed.Then,we propose a verification model based on the description model.First,we present an algorithm to reconstruct a task from subtasks' specifications.Then we present another algorithm to evaluate the risk of inconsistency.First,the acceptance function is used to compare the reconstructed task with original task.Then weights in specifications are used to compute the risk,so as to advise the requesters which task's non-functionality is the most suitable for improvement.This paper also includes the process management.Whenever a subtask is completed,the whole composition will be dynamically verified and adjusted.Firstly,reconstruct a task from subtasks.Then,evaluate the risk of consistency again and give the new suggestion.Finally,a number of static and dynamic cases have been designed and simulated to validate the effectiveness of the solution.The results show that the description model,reconstruction algorithm and the risk evaluation algorithm are correct and effective.It greatly reduces the burden on requesters and improves the accuracy,which meets the requesters' basic requirements.Also,the model and algorithm in this paper have advantages in scalability.They can meet the requesters' different requirements and also can meet the challenges brought by the future update of the algorithm.