Robust Word Sense Induction Shared Task at CoNLL-2025
Join the Robust Word Sense Induction shared task at CoNLL-2025, a unique opportunity to work with unsupervised learning and multi-annotated data in Natural Language Processing.
The Robust Word Sense Induction shared task is a part of CoNLL-2025, taking place in Vienna from 31.7 to 1.8.2025. This task focuses on unsupervised word sense induction without relying on predefined sense inventories. Participants will receive sentences containing target words and cluster them according to word sense usage. The evaluation approach is unique as it uses multi-annotated data and robust metrics that account for natural sense ambiguity, providing a fairer evaluation compared to traditional approaches.
The benchmark datasets will be available in English, Czech, German, Spanish and French. Important dates include 30. 1. 2025 – Trial data available
15. 2. 2025 – Test data available, evaluation starts
15. 4. 2025 – Test phase ends
22. 4. 2025 – Submission of system description papers
For more information and participation instructions, please visit https://projects.sketchengine.eu/conll2025/.
Tags: CoNLL-2025, Word Sense Induction, Unsupervised Learning, Natural Language Processing, Multi-annotated data, Robust Metrics