POLAR is a multilingual, multicultural, and interdisciplinary initiative focused on detecting and understanding attitude polarization in online discourse. From political debates and religious conflicts to ethnic tensions and gender bias, polarized language fuels misunderstanding, hate, and social fragmentation—especially in online spaces.
We created POLAR to address this growing problem by bringing together AI, linguistics, sociology, and ethics to develop responsible NLP systems that work across languages, events, and cultures.
Conventional NLP tasks often focus on sentiment or toxicity in a single language or event context. But real-world polarization is more complex:
POLAR fills this gap by designing resources, models, and benchmarks that can capture these nuances, offering a more realistic view of global online discourse.
POLAR is led by an interdisciplinary team of:
Our shared commitment is to open-source, ethically informed, and impact-driven research that bridges the gap between language and society.
Whether you're a researcher, developer, social scientist, activist, or student, you can contribute to the POLAR mission. Participate in our SemEval task, help us annotate new languages, or collaborate on interpretability methods.
Together, we can make language technology more inclusive, accountable, and socially aware.