File size: 761 Bytes
a1bceaf |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 |
**French-TV-transcript-NER** is a named-entity recognition model trained specifically on French TV headlines and transcript.
Given the format specificities, generalist multilingual or French model were unperforming. Additionally, the new model also provide additional set of entities useful in production (such as distinction between first name and last name).
## Entities
The model covers twelve entities:
* First name (prenom)
* Last name (nom)
* Location (lieu)
* Country (pays)
* Organization (organisation)
* Event (evenement)
* Nationality (nationalite)
* Broadcast name (emission)
* Product (produit), such as technological production, medicine, etc.
* Law (loi)
* Cultural creation (creation), such as movie titles, novels, etc.
* Disease (maladie)
|