This repository contains the full paper, code, link to models and ABSA guidelines associated with our paper.
We publish four fine-tuned versions of the NorBERT3 large model:
- norbert3-fine-absa: Sentence-level, fine-grained aspects, fine-tuned on the train split
- norbert3-coarse-absa: Sentence-level, coarse-grained aspects, fine-tuned on the train split
- norbert3-fine-absa-full: Sentence-level, fine-grained aspects, fine-tuned on the full dataset
- norbert3-coarse-absa-full: Sentence-level, coarse-grained aspects, fine-tuned on the full dataset
The two latter models are not evaluated, but represent what will be used in practice by NIPH/FHI (The Norwegian Institute of Public Health/Folkehelseinstituttet).
You can find the guidelines used for ABSA annotation in this repository. The pdf contains descriptions of each aspect, annotation examples and overall annotation guidelines.
Coming soon.