Before using the service, please read the preliminary information containing a description of steps that enable access to the CLARIN-PL developer interface.
Herference is used to detect coreference relationships in Polish texts. The result of the processing is saved in an HTML file containing a visualization of the coreferences that have been identified. The service uses the Herference tool developed at IPI PAN, implemented in the Clarin-PL infrastructure.
It uses the HerBERT3 model, a BERT model pre-trained for Polish text generation. The maximum length of the input data is 512 tokens, so longer texts are divided into pieces at sentence ends, if possible.
The service can be run:
No parameters.
The service can be run in the Windows system with default values using the following LPMN query: ['any2txt','herference']
[['any2txt','herference']] - input data in the form of a compressed directory (.zip)
Text.
An HTML file containing a visualization of the identified coreferences.
In Colab: Herference - Detection of coreference relations in the text
Karol Saputa (2022) "Coreference Resolution for Polish: Improvements within the CRAC 2022 Shared Task", Proceedings of the CRAC 2022 Shared Task on Multilingual Coreference Resolution, Association for Computational Linguistics: Gyeongju, Republic of Korea, 18–22.
(C) CLARIN-PL