Follow

Guide to: Quality Flow Project Set-up

 

This article is intended as a starter guide for users who want to learn how to set up or configure projects within the Quality Flow Projects framework, which is different from ADAP jobs and workflows. Please talk to your Appen contact if you'd like to know more.

Example Quality Flow Project

 

Overview

  1. Create your Project
  2. Upload data to your Project
  3. Create & Design your Work Job
  4. Configure your Work Job settings
  5. Route data into your Work Job
  6. Run your Job
  7. Create your Quality Assurance Job

Create a Project

Quality Flow can be enabled as a flag on your team, which you can request from your Appen contact. Once enabled, look for the “Projects” icon in the lefthand sidebar.

image-20230329-202137.png

When you click the icon, you’ll be brought to an empty project page, or a list of existing projects if someone else on your team has already created any, along with some high level information about the Project, such as the creation date, number of units and a raw progress bar.

image-20230510-030120.png

To create a project, click “Create Project”, in the top right of the Projects list page. The first pop-up you’ll see will look as follows. You will need to name your project and optionally give it a description. Once you click "Confirm" you will be taken back to the Projects list page.

 

image-20230329-234614.png

Once back in the Project list page, click on your new project; you will be taken to the DATASET page.

Quality Flow Navigation

On most screens you will see four tabs across the top of the page.

  • Dataset — this is your landing page where you upload and download data and perform actions on data such routing to jobs
  • Jobs — this is a canvas where you create Jobs and link them via flows
  • Curated Contributors — this is where your Appen Managed Services team will manage any curated contributors
  • Dashboards — this is where you monitor your throughput and quality, as well as download project-level reports

Upload data to your Project

On the DATASET page you'll be prompted to add data. You can skip ahead and create Jobs without adding data, but it is recommended to add data first as this will help to ensure your Job design and data structure are aligned from the start.

  • Quality Flow Projects expects a .csv or .tsv like other ADAP jobs
  • If you are using image, video or audio annotation tools, Quality Flow Projects expects input data to be an object such as a CDS link
  • If you wish to use the Unit Groups feature, please make sure to read this article on formatting your upload data first

 

You can add more data at any time by clicking on "Add More Data".

 

 

After you have uploaded some data, the table on the DATASET page will become populated. Additional columns will appear in the data table, such as the following, which are grouped into sections including “unit” and “source”.

    • Unit ID — this is an automatically generated ID specific to Quality Flow
    • Job title — this will be blank until the unit is sent to a Job, after that it will show the Job that the unit is currently in.
    • Job ID — this is the Job ID associated with the Job title.
    • Status — status of the unit. See Statuses.
    • Created At — this will show the date/time of the record.

To see all available columns, or to select which columns you want to view click on Columns.

  • Columns are filterable and sortable, but note that while your column selections will persist until you change them again, your filters will only persist while you do not navigate away from the page.
  • Used saved filters to manage your common filters

Notice the "Source" columns are deselected by default, select these columns to view your source data and check it has uploaded as expected.

 

 

Create a Job

To create a job in your project click on the JOBS tab. If you have already uploaded data, you’ll see a box on the canvas that says “ALL DATA” which will show the number of units you’ve uploaded. At the bottom of that box, click the small + to add a job.

If you have not already loaded data you will prompted to do so; it is recommended to add data before designing your job(s).

 

 

When you click the + at the bottom of the ALL DATA box, you’ll be prompted to give your new job a title.

Notice this job is called a "leading job"; Jobs that are created from the ALL DATA box are "leading jobs". For information on use cases & settings where there are multiple leading jobs, see this article.

 

image_20230522.png

Once you name your job and click “Confirm”, you’ll be brought to the template screen.

Notice that there are two boxes at the top: “Work Job” and “Quality Assurance (QA) Job”. You must select one of these. In most cases, you will start by creating a leading Work Job. After you have chosen "Work" choose "Start From Scratch" or else a job template.  See this article for information on usecases & settings for leading QA jobs.

image_20230522_1.png

 

Design your Job

From here, you will be taken to the DESIGN tab for the new job and you can proceed with designing your Job, as usual (see this article).

 

Configuring Leading Job Settings

In order to configure settings for your Work job, if you are not using Test Questions, and not ready to set up any manual review jobs, skip the QUALITY tab and go straight to SETTINGS, where you will see three settings areas: Test Question Settings, Rows Settings & Contributor Settings.

If you are using Test Questions, refer to this article, for Test Question set-up as well as information on Test Question specific Job Settings and Job Running instructions. If you are not using Test Questions move on to Rows & Contributors settings, described below.

 

Rows Settings

This is where you set your rows per page, lease expiry time and max number of units.

  • Judgments per row can be set from 1-200. Please see this article for important information about collecting multiple judgments in Quality Flow -- also note, the number of judgments cannot be changed once you have started data routing.

  • If you are using Unit Groups you will be asked to retain or ignore groupings, see Unit Groups for details.
  • Adjust the number of units you would like displayed per page.

  • Adjust the lease expiry in line with how much time you think a contributor will need to complete units. (If you are using Unit Groups, set the Assignment Lease Expiry time to allow the contributor to complete the entire unit group.)

  • Set the time limit alert threshold. For example if the contributor has 10 minutes to complete the page they are working on, and the alert threshold is set to 90%, the contributor will be alerted when they have 1 minute left to complete the page, and the timer will flash for the remaining time
  • If you want to limit the number of units a contributor can complete in the Job, ensure you select Enable max Judgments per Contributor and then specify the max number of units (this can be useful in early testing and callibration stages for example, and it can be adjusted after a job has been running). Otherwise, keep this toggle to OFF.

  • If you want to allow contributors to abandon (or skip over) units, you can select Allow Abandon Units (Abandoned units will go back to the contributor pool to be worked on by someone else, they will not go back to the original contributor). Otherwise keep this toggle OFF.

Save your settings.

 

Contributor Settings

External 

This is where your Managed Services team will manage any curated contributors. It is not yet possible for the general contributor population to work on jobs in Quality Flow Projects.

Internal

Here is also where you can assign contributors from your Project Resources (see this article), or copy the internal link to share with your in-house teams and start annotating.

Note that the Public link is disabled by default, you will need to enable it before sharing. You can change this setting after launch as long as the job is Paused.

Once you are happy with your selections, at the top of the page click on the button “Start Data Routing”. This will move the job from DRAFT to PAUSED state and make it available to have units routed to it. You can only aroute data into Jobs that are PAUSED or RUNNING see (see Job States). 

 

Sending data to a Job

Once you’ve instantiated your job, you’ll want to send data to it. After you click "Start data routing" you will be brought to the job's dataset page. Click "Go to Project Dataset" to start selecting the data you'd like to route to this job. See this article for data management options and suggestions). See this article if you are using Test Questions.

 

 

Once you have some units selected, click on the Actions dropdown, and choose Send Selected Units To A Job. Start typing the name of your target job in the field and select it when it pops up.

Note: Jobs in DRAFT state will not be available to send units to (see Job States).

 

 

From this pop-up you can also set your sample rate for routing:

  • if you want all of the units you’ve selected to go into this job, leave the sample rate at 100.
  • You can also choose the number of units you wish to send from this selection, or a percentage. This sample is random.
  • If you are sending units to a job to be worked on for the first time, you can ignore the options to carry over and overwrite judgments section, these options become important if you are sending units to be worked on again for rework or to collect additional judgments or annotations (see this article).

mceclip1.png

 

 

Refresh the page and the Data table will be updated. The units will now have been updated with their current Job Title and Job ID as well as a Status. You will also see the DATASET summary banner at the top of the page has become populated (see Unit Statuses for details).

 

 

The Job's DATA table will also be populated.

Actions you can perform from the Job’s DATA page are Assign/Unassign to Contributor. These actions are useful if someone has had units checked out and you want to give them to someone else to finish, or you want to assign particular units to particular people, in a practice job, for example.

image-20230403-222029.png

 

Run a Job

When you are ready to fully launch the Job to contributors, click on Run Job at the top right of the page.

 

 

Your Job will now show as RUNNING on the Canvas.

 

 

If you ever want to make the job unavailable to contributors, click on Pause Job at the top right of the Job's page (see Job States). You can restart at any time by Clicking on Run Job again later.

Leading Jobs that have never been used can be deleted, under certain conditions:

  • there are no units in the job
  • there is no submission history in the job

Any following jobs will automatically be deleted as well. To delete a job click on the Trash icon in the top right on any tab in the Job.

 

Create a Quality Assurance Job

Next - Guide to: Quality Assurance (QA) in Quality Flow Projects

 


Was this article helpful?
6 out of 7 found this helpful


Have more questions? Submit a request
Powered by Zendesk