Russian Nested Named Entity Recognition shared task: few-shot approach
Codalab with dev phase will be open even after the competition deadline. Codalab with test phase is open till 17 Feb. The max number of submits per day: 3.
Key dates
- 29 December: publication of the train set
- 4 February: publication of the test set.
- 25 February 2:00 pm (GTM+3): test phase deadline.
- 25 March 11:59 am (GTM+3): paper submission deadline.
Motivation
Named entity recognition is one of the most popular information extraction tasks in practice. It involves recognizing references of names, organizations, toponyms, and other entities in texts. The RuNNE competition is dedicated to the task of extracting nested named entities. Data markup allows the following cases: within one named entity there is another named entity. So, for example, the entity of the Organization class “Moscow Drama Theater of M. N. Ermolova” contains an entity of the Person type - “M. N. Ermolova”.
Data
The competition is held based on the material of the NEREL corpus, collected from the news texts of WikiNews in Russian. The NEREL corpus contains 29 classes of various entities, and the nesting depth of entities reaches 6 layers.
Shared task setup
We invite participants to consider a few-shot problem statement: the test set often contains named entities belonging to classes that are rarely found in the training set.