BRAT
The classic open-source web annotation tool behind a decade of NLP corpora.
About
BRAT, the brat rapid annotation tool, is the open-source web annotator from Stenetorp et al. (2012) that became the default choice for academic NLP annotation through the 2010s. It supports entities, relations, events, and attributes in a browser UI, with a client-server design for multi-user projects, and it remains widely cited. The GitHub mirror holds around 1,900 stars and 510 forks; code movement is slow, with the last push in mid-2024, but the tool is stable and complete rather than abandoned. One licensing note: the project is distributed as open source under the MIT license per its documentation, though the GitHub API reports no asserted license field, so confirm before embedding it commercially.
What it can do
- Web-based annotation of entities, relations, events, attributes
- Multi-user client-server projects
- Highly cited, battle-tested in academic corpora
- Lightweight standalone server
- Visualization of annotations in the browser
Setup
Clone the repo and run the standalone server; no build step required.
Prerequisites
Python 2/3 environment per the repo docs. Minimal resources.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (metrics tier, metrics-v1). Evidence: ~1,900 GitHub stars and 510 forks; Stenetorp et al. 2012 paper widely cited; stable codebase with 2024 activity. Strengths: de-facto academic standard for a decade, simple to deploy, enormous existing corpora in its format. Docked for: (1) slow maintenance cadence; (2) GitHub API shows no asserted license field despite MIT in project docs. License: MIT per project docs (GitHub API reports no asserted license).
★ 1,880 GitHub stars
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 9.0/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 8.0/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · Stenetorp et al. 2012; nlplab/brat with ~1,880 stars, 510 forks; last push mid-2024.
- · De-facto academic standard for NLP annotation through the 2010s; entities, relations, events, attributes.
- · License caveat: MIT per docs but GitHub API shows NOASSERTION; entry advises confirming before commercial embedding.
- · Manual review: stable and complete rather than abandoned; cited in the LightTag arXiv paper's related work.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.