To annotate your text training data on Diffgram, you first need to create a project. Here is a step-by-step guide to help you get started with your training data labeling on Diffgram. The process is standard and the same across all different data types viz Image, Video, 3-D, Audio as well.
What is a Project on Diffgram?
A Project is a top-level organizing object. A Project may have many datasets, users, task templates, labels, and more. A single project can handle multiple data types and can be shared to manage or work among different users.
Step 1: Creating a new project on Diffgram
To create a project, click on “New Project” under your current project tab as shown below and add the details of your project.
Step2: Creating labels and attributes
After creating the project, you would be creating both, Labels and Attributes.
Now go-ahead to create labels that you wish to annotate your text data with. For this example, we are trying to train a model to learn the difference between a noun and not-a-noun.
Attributes are how you wish to quantify/qualify the attributes of your labels. Diffgram has a variety of attribute tools ranging from free-select, free text, box to radio buttons, and more. Explore and settle on which setup will justify your attribution requirements. For everything else, add up a default value. Watch how it’s done in the video below.
Step 3: Uploading Data
Diffgram allows you to upload data from a cloud server or directly drop-upload from your desktop. Furthermore, you could add Pre-Labeled data (in JSON or CSV format) to the upload file.
Step 4: Inviting your team
Who is going to work on this project? Labelers, reviewers, and other team members can be added here along with the permissions you wish to provide them. You can skip this if you are a solo user and always come back to add more users.
Step 5: Task Management
The final step to launching your project is to set up task management. Here, you would create the list of tasks, people who are going to review these tasks, the percentage of tasks to be reviewed, which specific labels that they can or need to review, credentials required to be able to review OR annotate, set up the review pipeline and where the data can go after finishing up labeling and more.
Step 6: Annotate
You’re all set. Now go to project>studio-data explorer and get started annotating your training data. PS: This is how to add text annotations on Diffgram. For other data types, please refer to our annotation guides for Image, Video, 3-D and Audio.
Making A Project Public On Diffgram
You can make your project live to the public by going into the settings and clicking on “Make Public”. Once live, the project will be viewable by all users on the public internet. You can follow the same procedure to make your project private on Diffgram.
How To Delete Or Shut Down A Project On Diffgram
To delete a project on Diffgram, you need to submit a “shut down request”. To do this, you will have to go the settings.
Scrolling down, you will find a schedule for deletion option
You will also receive an email that notifies you of the deletion. Your project may be deleted in approximately 30 days unless you contact us to stop it.