Natural Language Understanding (NLU) is an area in machine learning that transforms human natural language into easily understood machine language.
The NLU team at AI Sweden has gained access to EuroHPC JU system VEGA in order to experiment with a Swedish GPT (Generative Pre-trained Transformer) model.
The access will be used to test the domain-specific capabilities of a Swedish GPT model using fine-tuning and p-tuning. Comparing these techniques will allow them to define the most resource-efficient way to adapt the model to specific use cases of the public sector. More concretely, in this research project, they want to fine-tune the model on conversational data to create a chatbot. They will also explore the more efficient technique of p-tuning to improve the conversational capacity of the model.