Doccano

Doccano is an open‑source annotation platform for natural language processing (NLP) tasks. It provides a web-based user interface that enables users to label text for a variety of supervised learning problems, including named entity recognition, text classification, sequence‑to‑sequence, sequence labeling, and language generation. The project was originally released in 2018 as a fork of the web application by the University of Texas at Austin, and it is maintained under the Apache License 2.0. Doccano supports multiple file formats—such as JSON, CSV, and XML—for both input and output, which allows easy integration with existing data pipelines.

The system offers role‑based access control, allowing administrators to create and manage projects, define label sets, and assign annotators with different permission levels. Annotation templates can be customized with conditional logic, making it easier to maintain consistency across large datasets. A real‑time visual feedback feature displays inter‑annotator agreement metrics, which helps teams identify discrepancies and refine guidelines.

Doccano is built with Django on the backend and uses the React framework for the front end, making it modular and extensible. It can be deployed on local machines, in Docker containers, or in cloud environments such as AWS, Azure, and GCP. The API exposes REST endpoints for programmatic interaction, enabling automated workflows for pre‑labeling, model evaluation, and annotation transfer.

The community contribution model is active; contributions appear in the GitHub repository under various issue trackers and pull requests. Regular updates add support for new data types, improved compatibility with transformer models, and integrations with tools such as Hugging Face’s datasets library. Doccano is widely adopted in both academia and industry for building high‑quality annotated corpora for machine learning projects.