The Appen Data Annotation Platform (ADAP) supports many different use cases, such as sentiment analysis, search relevance, data categorization, and image annotation. Before getting started on the ADAP platform, here are some tips to help create a successful job that will yield high quality results. For more information on getting started on the ADAP platform, please see these articles.
Choose a job template
-
Select the appropriate template that best fits the need of the project
- Check out the templates or start from scratch
- Preview the templates to see the layout of each job design and instructions
- For more information on ideal jobs and common use cases to run on the platform check out this article
Add Data
-
It is important to format the data before uploading to the platform.
- It is best to use a UTF-8 encoded .CSV, however .TSV, .XLSX, and .ODS file formats are also acceptable and require UTF-8 encoding.
-
All data files should:
- Have a unique column header for each column.
-
Be uploaded to the Data page.
- Note: unless updating existing test questions, these should be uploaded to the Quality page.
- To view more related articles, click here.
Design
-
Graphical Editor
-
Functions:
- Create or modify job title
- Insert Data
- Generate questions without typing any code.
- Best used for those who are not familiar with CML.
- Can be used to customize templates or validators to ensure inputs are entered correctly, and use logic to hide or show certain questions.
- To view more related articles, click here.
-
Functions:
-
Code Editor
- The Code Editor is used to build custom jobs with HTML, CSS, Javascript, and Custom markup language, CML.
- It can perform all functions as the graphical editor and more.
- To view more related articles, click here.
-
Instructions
- Effective instructions in the job will help the contributors understand and correctly complete the task.
- They should cover the steps to complete the job, as well as helpful tips and examples.
Quality
-
Test Questions are:
- The most important quality control mechanism in the ADAP Platform.
- Rows that are pre-labeled with known answers that are randomly inserted throughout the job.
- Will be used to test and track contributor performance.
- Ensure that only the contributors demonstrating competency in the job are allowed to submit judgments.
- To view more related articles, click here.
Settings & Launch
-
The Launch page contains settings for:
- Judgment per row
- Payment per row
- The number of rows that are going to be launched for that particular run.
- Definitions and other common settings can be found here.
Monitor
-
Job monitoring is helpful to:
- Check on the health of the job
- Identify areas for improvement.
- Reviewing the dashboard provides a high-level overview of the job.
- By far the most important aspect of monitoring the job is reviewing the test questions.
- To view more related articles, click here.
Results
-
After the job has completed here are some next steps for to view the results:
- Edit report settings
- View the results
-
Audit the output data
- Note: The audit is extremely important for iterating on the job, as it helps find areas of opportunity for improving contributors’ understanding of the task.
- To view more related articles, click here.