Milestone 3 sanjosespartans PowerIdea 2: Workflow Planning as Decentralized Optimization Problem
Preparing complex jobs for crowdsourcing marketplaces requires careful attention to workflow design, the process of decomposing jobs into multiple tasks, which are solved by multiple workers.
While workers decompose and solve tasks, requesters can view the status of worker-designed workflows in real time; intervene to change tasks and solutions; and request new solutions to subtasks from the crowd.
Crowdsourcing marketplaces are increasingly used to solve computational problems that require human intelligence,such as audio transcription and data verification.
Many problems on markets like Amazon Mechanical Turk are solved as a series of microtasks: narrowly focused, brief tasks designed to be completed in a few minutes, such as labeling an image or checking the accuracy of data through web search. For quality assurance and to accomplish complex work, multiple microtasks are frequently chained together into workflows. We can break down large complex tasks into smaller ones and then proceed with focussing on the smaller ones and then moving ahead towards the remaining pieces and then solving them accordingly.
A good WorkFlow Design is a challenge till date and involves substantial planning,software development and testing. Requesters generally rely on iterative process to construct good workflows. Requesters guess a viable workflow ,implement all of its steps as a software that interfaces with a crowd platform,test it with live workers ,identify points of failures,bottlenecks encountered if any and then resolve the problems and then acccordingly modify the workflow.
The responsibility of workflow design can be shared between the requester and the worker.
Effective technique can include a workflow manageemnt system that aims at first breaking down the complex tasks into smaller ones as in the cse of Divide and Rule problem criteria. Requesters are able to delete or modify plans that are made by workers,request the crowd to replan components of workflows or seed the system with aprtial plans to evaluate their effectiveness.
Early work emphasized on tool for processing large Data Sets in applications like taagging and classification that were outside the reach of autonomous algorithms. AI Researchers have emphasized the utility of crowds for supporting active learning programs. While most crowdsourcing tasks are designed manually based on prior experience and intuition, formal techniques can also play a useful role. With the introduction of new techniques like Map Reduce they can be effectively applied to large sets of data which we get from these applications and then analyze the data and then devise suitable methods based on it.
Algorithmic Model :
It needs to be given a thought at the current level and then suitably we can proceed in this direction.As of now we have thought of a plan like Working on the lines of Divide And Rule Paradigm and after that Effective Merging Techniques wherein each divided work can then be joined with the similar task and then combined together to give back the original Problem,include Worker,Requester and Crowd Interaction on them . Make use of Map-reduce technology when there would be multi data sets to be read from the system and make use of effective Daat Analysis tools that in turn would help in identifying suitable trends in the system and then would help in refining the User Experience and the User Workflow.
Crowd Planning and Execution :
To know how effectively crowds can be used to support the execution of complex work, two evaluations can be performed they are : 1.Unsupervised Crowd Planning : Here we allow the Crowd to perform the changes without any expert intervention 2. Collaborative Planning and Execution : It includes to recruit more experienced workers and another strategy is to allow requesters to monitor and selectively intervene in the workflow design and execution process.
End Result :
As our present system of Crowdsourcing from Stanford is going,we are been given tasks where work is done by the crowd and each team has members having people with different skills ,so the end result is that this might end up being a CrowdSourcing Platform where there might be better interaction between the Requester , Workers and it might overcome the previous Bottlenecks of the Crowd Sourcing Platfroms,that have been mentioned in the past.