metadata
license: apache-2.0
language:
- it
metrics:
- accuracy
library_name: bertopic
pipeline_tag: text-classification
tags:
- legal
- text-generation-inference
Fine-tuned BERT model for Italian laws domain classification
Each Italian law can be classified according to their domain, which refers to the minister(s) responsible for signing the law. The model, by looking at the title of the law, classify the law into one of the domains that are a direct derivation of ministers. The possible labels are:
| Ministry | label |
|---|---|
| interno | 7 |
| istruzione | 9 |
| presidenza | 11 |
| esteri | 5 |
| difesa | 3 |
| pubblica amministrazione | 12 |
| economia | 4 |
| trasporti | 15 |
| giustizia | 6 |
| lavoro | 10 |
| agricoltura | 0 |
| comunicazioni | 1 |
| sanita | 13 |
| spettacolo,sport,turismo | 14 |
| cultura,ambiente | 2 |
| istituzioni | 8 |
Performances of the model
- Average training loss: 0.23
- Accuracy on a validation set: 0.90