transcript-text-analysis is an encoder model specialized for the classification of French news transcripts. The model is based on debertav3 and has been trained on 1,018 examples of annotated transcripts.
Given a text, transcript-text-analysis will generate the following classifications in French:
- Emotion (Neutre, Persuasif, Optimiste, Solennel, Alarmant, Indigné)
- Expression (Interview/Discussion, Publicite, Informations, Meteo, Reportage/Enquete)
- Intention (Informer, Sensibiliser, Promouvoir, Mobiliser, Divertir, Eduquer)
- Theme (Santé, Société, Économie, Politique, Sports)
- Tonalite (Informative, Emotionnelle, Publicitaire)