ALToolbox

What it does

Annotation interface

It provides a serverless GUI annotation tool integrated into Jupyter.checked 14 Nov 2026

Data reusability

It includes a PLASM postprocessing pipeline for annotated-data reusability.checked 14 Nov 2026

Datasets

Custom datasets can be loaded from Hugging Face datasets or repository data files.checked 14 Nov 2026

Installation

The README instructs installation with pip install acleto.checked 14 Nov 2026

Jupyter setup

The annotation widget requires additional installation steps for Jupyter Notebook or JupyterLab.checked 14 Nov 2026

License

The project is licensed under the MIT License.checked 14 Nov 2026

Model compatibility

The repository supports Hugging Face Transformers models for sequence and token classification.checked 14 Nov 2026

Model distillation

It includes acquisition-model distillation and domain adaptation tools.checked 14 Nov 2026

Purpose

ALToolbox is a framework for active-learning annotation in natural language processing.checked 14 Nov 2026

Query strategies

It includes strategies such as ALPS, BADGE, BAIT, BALD, BatchBALD, and Entropy.checked 14 Nov 2026

Research audience

The publication describes the framework as intended to support data scientists and researchers.checked 14 Nov 2026

Supported tasks

It supports text classification and sequence tagging tasks.checked 14 Nov 2026

Testing

Tests default to CUDA when available and otherwise use CPU; CPU is recommended for robust results.checked 14 Nov 2026