A simple web application to review and validate Named Entity Recognition (NER) results from AI pipelines. Perfect for quality control of automatically extracted entities like people, organizations, and locations from article texts.
This validator is designed to work with outputs from the IWAC AI NER Pipeline, which extracts named entities from articles using AI models. Use the pipeline to extract entities, then use this validator to review and correct the results.
- Review articles with their AI-extracted entities side by side
- Validate or reject each entity with a single click
- Add missing entities manually
- Track your progress across multiple sessions
- Export clean CSV files with only the validated entities
-
Make sure you have Docker installed on your computer
-
Download or clone this project, then open a terminal in the project folder
-
Start the application:
docker-compose up --build
-
Open http://localhost:8081 in your browser
-
Upload your CSV file and start validating!
- On the home page, drag and drop a CSV file or click to browse
- Your file should have these columns:
o:id,Title,bibo:content,Subject AI,Spatial AI - Click "Resume" on any session to continue where you left off
The screen is split into two parts:
- Left side: The article text with entities highlighted in color
- Right side: List of all entities to validate (resizable - drag the edge!)
- Click the checkbox to validate an entity (click again to undo)
- Click ✓ to validate
- Click ✗ to reject
- "Validate All" approves all pending entities at once
- "+ Add" lets you add a missing entity manually
Entities are sorted alphabetically to make them easier to find.
- Hover over any entity in the right panel to highlight it in purple in the article text
- Yellow highlights = Subject entities (people, organizations, topics)
- Green highlights = Location entities (places, countries, cities)
- Use the search bar to find specific words (highlighted in blue)
- Use Previous/Next buttons or arrow keys to move between articles
- Skip moves to the next article without saving
- Save & Next marks the article as validated and moves on
Click "Export Validated CSV" to download your results. The exported file will:
- Include only validated entities
- Remove all rejected entities
- Include any entities you added manually
- Keep the original article data intact
Your CSV file should have these columns:
| Column | What it contains |
|---|---|
o:id |
Unique article ID |
Title |
Article title |
bibo:content |
Full article text |
Subject AI |
People, organizations, topics (separated by |) |
Spatial AI |
Locations (separated by |) |
- The panel on the right can be resized by dragging its left edge
- Hover over an entity to quickly find it in the article text
- Use keyboard arrows for faster navigation between articles
- Your progress is automatically saved - close and come back anytime
Port already in use? Change 8081 to another port number in docker-compose.yml
Changes not appearing? Run docker-compose down then docker-compose up --build
MIT