Advanced Controls

For our power users that want more control in terms of pipeline setup, labeler pool characteristics, and other advanced features can leverage our advanced controls.

This configuration allows for you to choose the number of reviewers for a task and whether you want to auto redo rejected audits.

This configuration allows for you to choose the number of reviewers for a task and whether you want to auto redo rejected audits.

On the point of pipeline configurations, Rapid actually supports several pipelines.

Traditionally, CV tasks usually go through the Tasker then reviewer pipeline where a Tasker attempts the task followed by a number of reviewers.

For NLP tasks, we offer the reviewer pipeline, consensus pipeline, and freeform pipeline. The consensus pipeline allows for you to leverage multiple Taskers and their votes to generate the final answer. The freeform pipeline is designed for tasks that can have many right answers.

For Audio Transcription, we also have another layer of abstraction that you can indicate when you create your project as shown below.

You can select the pipeline type and whether it is related to audio transcription or not in the project creation page.

You can select the pipeline type and whether it is related to audio transcription or not in the project creation page.

Next is the labeler pool characteristics. We allow you to select the relevant languages. Note that this will resulted in a higher surcharge depending on the language but you can always estimate using the calculator on the lower left hand side.

Notice that I selected Basic Japanese as the requirement for my labeler pool.

Notice that I selected Basic Japanese as the requirement for my labeler pool.

Lastly, there are a few more advanced controls.

Edge case alerts allow for Taskers to surface edge cases to you. Common errors allow for automatically converting rejected audits into common errors. And common confusions are used to distinguish between confusing categories.

Edge case alerts allow for Taskers to surface edge cases to you. Common errors allow for automatically converting rejected audits into common errors. And common confusions are used to distinguish between confusing categories.

Updated a month ago