![DalTagArchitecture](https://user-images.githubusercontent.com/51812121/221656208-10535d41-938e-4d5b-a08a-faef15e00b6b.png)
### DALTag: An Efficient Text Annotation Platform Supporting Few-shot and Active Learning

We propose DALTag, an on-demand deep active learning-based text annotation platform that leverages few-shot learning (FSL).DALTag uses different active learning acquisition techniques that rank the unlabeled data points according to the model's confidence in their automatic annotations. It identifies the top data points with initial annotations. Users can decide whether to keep the annotations or update them. The updated annotations are used to fine-tune the pre-trained models using FSL. Users can easily configure DALTag based on one of its built-in models and techniques. Moreover, DALTag allows users to add their own acquisition techniques and datasets, enabling efficient handling of new datasets and text classification tasks. During the demonstration, the attendees will witness the ability of DALTag to annotate data samples using active learning and a few human-annotated samples. We evaluated DALTag using different text classification tasks, e.g. sentiment analysis and named entity recognition (NER).DALTag achieves results equivalent to training using the whole training data by using 45\% of training data in the Question Classification task and 30\% of training data in the Named Entity Recognition task, making it a flexible and accurate solution for annotating text data. Our DALTag demo video is available online: https://rebrand.ly/daltag

### Demo: DALTag mode and its three main functionalities 

![image (1)](https://user-images.githubusercontent.com/51812121/221657331-8dfddda5-4033-46e1-9841-6c74c1c7675a.png)

### Features:
1. Model Training with Few-shot learning
2. Active Learning based sample selection
3. Auto-labelling for assistance

### How to run DALTag
    pip install -r requirements.txt

## Run the backend (Python 3.8) 
    cd backend
    python manage.py migrate
    python manage.py create_roles
    python manage.py createsuperuser
    python manage.py runserver

## Run in a separate terminal 
    cd backend    
    celery --app=config worker --loglevel=INFO --concurrency=1

## Run the frontend
    cd frontend
    yarn install (use node version 16.0.0 && yarn version 1.22.19)
    yarn dev