Connect now supports 2 new project templates, mainly geared for AI training applications:

• Data Labeling/Media Annotation
• Custom HTML

What is Data Labeling?

At its core, data labeling involves the task of attaching meaningful information to raw data, predominantly for the purpose of training machine learning or artificial intelligence models. The process entails annotating media, such as images, videos, audio, or text with relevant labels to aid AI systems in understanding and processing the content effectively. This method, therefore, stands as a cornerstone for many AI applications, enabling them to function accurately.

How Are Data Labeling Templates Different From Surveys:

1. When conducting surveys or experiments, it is critical that each respondent only participate once. Thus, the Connect survey system prevents respondents from participating in a survey more than one time.

For AI training tasks, on the other hand, respondents can complete as many tasks as they want. Therefore, there is no restriction on the number of labels or responses a participant can provide.

2. When conducting surveys or experiments, a third-party platform such as Qualtrics or Survey Monkey is used for stimulus presentation. A survey link that participants use to access the survey always accompanies a Connect survey project.

For AI training tasks, on the other hand, stimuli are uploaded from a csv file or using HTML. The data labeling templates are, in essence, their own powerful type of study creation software. They enable a researcher to provide instructions and to display images, videos, audio,  or text stimuli that participants can respond to.

Below, we provide more detail on how to work with the Data Labeling Templates on Connect.

Choosing a Template Type

There is no need to create your desired project with a 3rd party platform as these types of projects will be executed within Connect itself.

Choose your desired template and click on "Next" to continue the setup. You can set a default which will load your selected template for any subsequent projects created.

Creating a Data Labeling Project

The process of creating a project remains generally the same as with survey based projects on Connect. You can read about that here: https://researcher-connect.zendesk.com/hc/en-us/articles/4416207746068

With regards to this new Data Labeling template, we have additional fields that require your attention:

Upload Your Data

Whereas survey based projects require you to specify how many participants you need to target, these new templates require you to specify how many tasks you need completed. 

To do so, please prepare a CSV (comma separated value) file with the following format:

• Row 1 - Header Name
• Row 2 - data that will be shown to the participant(s)
• Row 3 - data that will be shown to the participant(s)
• .......
• Row N - data that will be shown to the participant(s) (where N is the number of total tasks you are uploading)

Each row represents a task and contains links to media files that you have hosted elsewhere. Formats that we support are:

• Audio
• Image
• Video

Connect will recognize the file type and the number of rows in the file you uploaded. Please review and confirm before proceeding. Please note you won't be able to proceed if there are formatting issues in the file.

Participant Tasks, Time and Cost

You will now have the ability to allow participants to take more than 1 available task in your project. You can specify whether there is a maximum or unlimited number of tasks they can take.

Since tasks that require data labeling or media annotation, especially in the area of AI training, are relatively quick, we now support the ability to set a task to be completed less than a minute. Please note that our payment minimums still apply for these microtasks.

Prior to saving and launching the project, please make sure to preview the task to ensure that the instructions you provided are clear and that the media you specified are enabled and working.

Launch and Collect Your Data

Simply launch the project as you would normally. Data that the participants are submitting will be part of the CSV file you download from the "Participants" table within the "Manage Participants" page:

Please note that the data will be encoded in a raw format and additional processing will be required to extract/clean the submissions.

Creating a Custom HTML Project

Custom HTML template projects are similar to our Data Labeling projects with the exception that you can control how you want tasks to be displayed to the participant. This template requires knowledge of HTML and programming concepts, so this may be more suited for power users.

Upload Your Data

While you are still required to upload your tasks in a CSV (comma separated value) file, you add all the columns you require to be visible on your project. Each column in the CSV file will become a variable that you can inject into your HTML design.

For instance, having a column with the header "Animals" will allow you to set that as a variable called

task.row_data['Animals'] in your project design:

 

Input fields that you do collect require that you use a unique NAME attribute, so that each and every piece of data you collect is identifiable in your download:

Please refer to the "Data Label Projects" section above for notes on tasks, timing, launch, and cost details.