Model Description
This model builds upon the pre-trained NB-BERT model developed by Nasjonalbibloteket, which is designed to handle the nuances of the Norwegian language. The fine-tuning process tailored the model to understand and classify the sentiment of Norwegian governmental communications, particularly during the COVID-19 pandemic
Use cases
Sentiment Analysis of Governmental Communication: Objective: Analyze and classify the sentiment of official governmental communications from Norwegian authorities. Example: Categorizing tweets from the Norwegian health department and government announcements to understand public sentiment trends.
Limitations
The model is fine-tuned specifically for Norwegian
The model is only trained for government specific sentiment analysis
Training data
Source: The datasets used include: Twitter data from the Norwegian Health Department (Folkehelseinstituttet) and the government (Rejeringen). A timeline dataset of the Norwegian government's handling of the COVID-19 pandemic. The labeling process utilized a combination of Chatgpt4 and manuel corrections, the fine-tuning is included
Credit
This work builds upon the NB-BERT model. For more information on the original NB-BERT model, please refer to the Nasjonalbibloteket community’s publication: https://github.com/NBAiLab/notram