ScaleHub files patent application for Dynamic Crowd Balancer
We are excited to announce that we have filed a patent application for a new innovation that dramatically improves the ScaleHub Portal. Our Dynamic Crowd Balancer uses AI to automatically distribute work across multiple crowdsourcing marketplaces simultaneously. Equipping the ScaleHub Portal with the Crowd Balancer is instrumental to meeting service level agreements—both our own and those of our customers.
How it works
The patent-pending Crowd Balancer manages the allocation of tasks across multiple micro job platforms like Amazon Mechanical Turk (mTurk), for example. Users can view and manipulate workflows at a glance across all platforms while AI-based algorithms dynamically change workflows to complete micro jobs more cost effectively.
To better understand how everything works, let’s use the example of medical record transcriptions. Imagine you have 100 medical records to transcribe, a task that you will distribute in stages between crowdsourcing marketplace A and crowdsourcing marketplace B.
ScaleHub users have the ability to break larger jobs into more discrete tasks. In this example, you could use ScaleHub to break each medical record into “snippets”—smaller subsections of a larger record. Let’s suppose each of the 100 records is broken down into 10 snippets. You now have 1,000 micro jobs. Crowd contributors complete micro jobs in less time than full records, and multiple contributors can take them on simultaneously.
Our new technology improves this workflow by:
1. Dynamically balancing load of micro jobs across platforms
What happens if one platform is completing tasks at a higher rate than the others? If you add tasks in stages—let’s say 100 snippets at a time, 50 going to each platform—our algorithm monitors the progress for each, and distributes future tasks accordingly. If the workers on platform A complete all 50 while those on platform B still have 20 left in their queue, the next round of micro jobs gets assigned in such a way as to equalize the workload. In this example, a new batch of 100 jobs is split up with an eye toward balance: 60 new jobs are routed to platform A and 40 go to platform B so that each platform will have 60 in total left to finish. This process continues until an entire queue of tasks is completed, with users submitting micro jobs en masse and the algorithm splitting them up according to throughput.
2. Monitoring centralized tasks
The addition of the Crowd Balancer gives users a more refined view into their workloads and progress through a single landing page. As each of our transcription tasks is assigned, balanced and completed, users can see the real-time status for tasks—or group of tasks—on each platform. Insights into how many tasks are completed, how many remain, the time left to completion and how many workers are available are easily understood at a glance.
3. Automatically reassigning tasks
As a final improvement, the Crowd Balancer can automatically reassign work away from platforms experiencing performance issues. If any of the available crowdsourcing platforms suffer a server or network outage, the Crowd Balancer will route remaining tasks away from the problem site or reassign them to sites running smoothly. Additionally, we can detect the circumstances that led to past outages and redirect tasks toward higher performing platforms.
These improvements to our already leading-edge technology represent a giant step forward for ScaleHub—precisely why we’re so happy to share this news with you.
Read about the magic behind our crowdsourcing solutions, or check out the ways innovative businesses are putting them to use.