doccano

Why we included this project

Most open source models only perform as well as the labeled data behind them, and doccano exists to help you build that data yourself. It is a self-hosted web app for text annotation that covers the tasks practitioners actually repeat: text classification, sequence labeling like named entity recognition, and sequence-to-sequence work such as summarization. You create a project, upload documents, define guidelines, and coordinate labels across collaborators, with multi-language and mobile support included. Installation is straightforward via pip, Docker, or Docker Compose, and a REST API lets it slot into an existing labeling pipeline without forcing a proprietary SaaS subscription. For a small team or solo practitioner assembling a domain-specific training set, this is a practical way to produce labels you own and control.