Website

M2QA Data Annotation Platform

Tutorial

One crucial part of creating a high-quality benchmark dataset is a good tutorial. This tutorial introduces the crowdworkers to the task and teaches how to provide high-quality data. Below (in the "Screenshots From the Tutorial" part), you can find our tutorial, or you can have a look at the source code here: app/templates/tutorial

Installation

1. Create conda environment: conda env create --file environment.yaml
2. Activate conda environment: conda activate m2qa
3. To install all dependencies for the frontend, follow the instructions from app/README.md
4. Create a system service under /etc/systemd/system/mmqa-website.service with this content: mmqa-website.service
5. Install nginx and add this file nginx-config-file.txt to /etc/nginx/sites-enabled/mmqa
6. The website is password protected. Generate admin and user passwords:
   • sudo htpasswd -c /etc/nginx/admins.htpasswd <ADMIN_USERNAME>
   • sudo htpasswd -c /etc/nginx/.htpasswd <NORMAL_USER_USERNAME>
7. If you want to use DeepL to translate the texts during your quality control, add your DeepL authentication key in plain text in Website/app/deepl_auth_key.txt

Run the Annotation

Import Data to Database

flask db import ../data/de/news/articles.json

Add Prolific Study

flask db add-study <study_id> <prolific_key>

For other commands, have a look here: app/cli_db.py and app/cli_statistics.py

URL

• Needs PROLIFIC_PID, STUDY_ID, SESSION_ID and language: <URL>?PROLIFIC_PID=<PROLIFIC_PID>&STUDY_ID=<STUDY_ID>&SESSION_ID=<SESSION_ID>&language=<LANGUAGE>

Screenshots From the Tutorial

Upon beginning an annotation session, the crowd-workers see this Overview page:

They then go through the tutorial, in which they have learned to use our interface and are introduced to the task.

Tutorial Step 1

Tutorial Step 2

Tutorial Step 3