This repository contains the pre-trained models for Named Entity Recognition and Linking for historical Italian presented in the paper **"DELICATE: Diachronic Entity LInking using Classes and Temporal Evidence"**.

The directory contains the following folders:

- **bert-base-italian-xxl-uncased**: containing the pre-trained BERT model used by the bi-encoder
- **blink_biencoder_base_wikipedia_ita**: containing a BLINK bi-encoder trained on the Italian Wikipedia
- **GBT**: containing the Gradient-boosted Trees re-rankers trained on the [ENEIDE](https://github.com/sntcristian/ENEIDE) dataset
- **NER**: containing GliNER-large models trained on ENEIDE
- **KB**: containing the faiss Index for BLINK and a Sqlite database containing metadata about the entities

Documentation for using the following models can be found in the [DELICATE](https://github.com/sntcristian/DELICATE) repository.

---
license: gpl
---