🤖 A Text Annotation GUI powered by machine learning 🤖
TAG is a web-based text annotator that allows you to upload documents and highlight them using custom labels and highlight colors.
Oh, and there's a machine that can annotate and categorize the words in the document for you.
Use it here: http://tagweb.pythonanywhere.com
- Modern browser (Chrome/FireFox)
- Desktop for best experience
- Annotate entire document(s) using our spaCy model
- Or take the annotating into your own hands and highlight manually
- Customizable label names and label colors
- Highlight/unhighlight using searchbar (supports both regular search AND regex)
- View all current highlights by hovering over the right-side panel
- Click to scroll to that specific highlight
- .txt and .dox file upload support
- Also supports .zip containing file types of the above
- Download your results!
- Options include a single JSON of the current document containing the document text, the annotations, and their corresponding labels
- Extra option to download a .zip of JSONs if multiple documents have been uploaded
- Upload a document (or multiple) by clicking the '+' next to the 'Documents' section on the left panel
- Then, you can have the spaCy model annotate the document for you using their default categories by hovering over the 'Annotate' button found at the bottom and choosing an option
- If, on the other hand, you choose to use TAG manually, you can start by clicking the '+' next to 'Labels'
- From here, you will be prompted to enter a name for this category
- By default, a highlight color will be generated at random
- This can occasionally lead to an ugly color, in which case you can change the highlight color by clicking on the eyedropper and then selecting your preference
- Finally, you can start highlighting
- Double click or just click and drag to highlight
- If you want to highlight a word again OR correct it, you can highlight again and a menu will prompt you as to whether you want to add a new highlight or delete the highlighted portion of the existing highlight
- For shorthand usage, we added the ability to add an annotation that is already an annotation by holding down the 'a' key while highlighting
- Likewise, for deleting, hold down 'd'
- If you want to highlight by search, use the designated search bar
- 'txt' denotes regular searching, clicking it will change to 're', denoting regex searching
- The '>' below the search bar is a switch for the context of the search
- 'Add' will highlight all of the instances of the searched word(s) while 'Delete' will do the very opposite
- View the progress of your annotations by hovering over the right panel
- Click an annotation to scroll to its position in the document
- When finished annotating, you can download the condensed results as a JSON by hovering over 'Download' in the bottom left of the screen
- Options include a .zip of all the open documents or just a singular .json of the current document
- Training
- To create your own offline spaCy models and train them on data sets, see our related open-source project, TAG-Local
Contributions are what make the open source community such an amazing place to be learn, inspire, and create. Any contributions you make are greatly appreciated.
- Fork the Project
- Create your Feature Branch (
git checkout -b feature/AmazingFeature) - Commit your Changes (
git commit -m 'Add some AmazingFeature') - Push to the Branch (
git push origin feature/AmazingFeature) - Open a Pull Request
Distributed under the MIT License. See LICENSE for more information.
- Matthew Anthony
- Nicholas Gonzalez
- Keith Hudock
- Tony Mark
- Leo Vergnetti
- Qunchao Zhou