ActiveTigger: A Collaborative Text Annotation Research Tool for Comp. Social Sciences
summary
ActiveTigger is a working open source annotation stack: a Python backend and React frontend that computes embeddings, offers a collaborative (asynchronous) annotation interface, runs an active learning loop to pick items, fine tunes BERT-like encoders (400 MB-ish bases on lab GPUs) and reports metrics like F1 for a held-out test set. It already annotates tens of thousands of texts via export and inference, supports on‑prem GPU deployment, topic modeling, LLM-as-a-service queries for scaling, and a Python client for automation.
AI-written summary · about
chapters (9)
- 0:00ActiveTigger introduction
- 3:15Need for annotation tools
- 6:25ActiveTigger origins
- 8:33Goals behind ActiveTigger
- 10:41Technical architecture
- 12:06Live demo
- 19:21Future features
- 21:29Conclusion
- 22:12Q&A
related talks
37mBuilding AI-powered user interfaces
10mSGToolkit: An Interactive Gesture Authoring Toolkit for Embodied Conversational Agents
12mBuilding a TinyMCE LMS Editor with React & OpenAI - Pt 1
01h 05mBuilding a TinyMCE LMS Editor with React & OpenAI - Pt 2
02mDBTB INT Aaron Kalb r
42mAI in React JS
24mSupercharging Dev Workflows: AutoGPT Meets React Flow
15mIncremental Game Mechanics Applied to Text Annotation