Ads Area

HITs and Assignments in Mturk - The best work from home online job in india

If you reached onto this article by reading my previous article, then start reading this article ASAP. If not, you should read my first article to get the basics. It's not an option, but a must need to understand this article.
Amazon Mechanical Turk (MTurk) is a powerful vehicle for getting tasks done quickly and efficiently but it works a bit differently from other crowdsourcing marketplaces. The terms HIT and Assignment can be confusing to those that are new to MTurk and just want to get a spreadsheet worth of data processed. In this tutorial we’ll explain how to go from a spreadsheet to completed results and provide details on how we made decisions about HITs and Assignments.

Starting from a spreadsheet

Many Requesters start using MTurk with a spreadsheet they want processed. One option is to send this big file of data off to a freelancer and ask them to spend the next week on the task. While that is a good option in some cases, it typically takes far too long and there aren’t great ways to measure the quality of the work being done by your freelancer.
The power of MTurk is in the ability to slice that spreadsheet up into individual tasks that can be completed by many Workers. By splitting it up, you get results much more quickly. And by asking that each task be done by two, three, or more Workers, you can ensure that the results you get back are of the highest quality possible.
Let’s start with an example, we want Workers to identify the objects in all of the images stored in a spreadsheet:
The easiest way to approach this is to view each row as a task you want completed. In MTurk, this is a Human Intelligence Task (HIT). Each HIT will ask a Worker to identify the object in the image. A Worker might do just one of these HITs, or dozens. By allowing hundreds of Workers to work on your project at once, you get faster results.
Faster results are only part of the story. With MTurk it’s easy to assign the same HIT to multiple distinct Workers. These are called Assignments and it’s a powerful way to increase the accuracy of your results. In this case, we’re going to assign each item to three Workers:
Even the most diligent Worker will make a mistake occasionally so assigning HITs to multiple Workers is a great way to check for response quality. If a Worker makes a mistake and two other Workers agree on the correct answer, you can use their result as an answer.
For example, here Worker A2XXXXXXXX accidentally clicked the wrong button and selected rabbit instead of dog, but because we have two other Workers letting us know it’s a dog, we can assume their answer is correct. If for some reason all three Workers disagreed, that could be a signal that the image or the instructions are confusing and it needs to be reviewed separately.

MTurk in practice

Let’s set this Project up in MTurk. To start we’ll visit the Create tab on the Requester website and select Image Classification and Create Project.
In your HIT Properties you can specify information about your project including a description, reward amount, and how many Workers you want assigned to each HIT. Here we are assigning the HIT to three Workers and expect that total work time across all three Workers will be about 18 seconds, or 6 seconds per Worker per HIT.
Now we can continue on to design the layout of our task, including adding instructions for our Workers. More information on doing this for various task types including image categorization can be found in other MTurk Tutorials.
Now that we have our Project defined, we can publish the data in our spreadsheet for Workers to start working on. We recommend starting small so that you can make sure everything is working correctly. First, we’ll copy a dozen rows from your spreadsheet to a new Workbook and save this data in comma delimited (CSV) format.
We can now select Publish Batch to load this CSV file and publish it to Workers.
Workers will find and complete these HITs.
When we return to the batch management page, we see that all of the Assignments have been completed and can now download the results.
The results will typically show Workers generally agreeing on answers with occasional cases of disagreement:

Next steps

There are a variety of techniques for identifying and handling cases where Workers disagree depending on the type of task being performed. We’ll address those in future tutorials.
We hope you found this to be a helpful introduction to HITs and Assignments. If you have any questions, please post a question to our MTurk forums. 
So, now you are perfect to Join Amazon's Mturk, use the below ink to sign up and start working. Click the link below and go and create a Creator Account and follow the instructions. (Remember, you will have spend some time to be familiar with the website and know the features and settings it has. No one can help you as a guide through it, you will have to practice it on your own; Once you mastered the website, then you will earn every second you set your mind to). Now go and sign up!

Post a Comment

* Please Don't Spam Here. All the Comments are Reviewed by Admin.

Top Post Ad

ശ്രദ്ധിക്കുക : നിങ്ങൾ കുവൈത്ത് പ്രവാസി ആണെങ്കിൽ, ബൂസ്റ്റർ ഡോസ് ഉടൻ എടുക്കുക. ഓൺലൈനായി എങ്ങനെ രജിസ്റ്റർ ചെയ്യണം എന്ന് ഇവിടെ വിശദീകരിക്കുന്നു