Objectives
- Select suitable AI model
- Creating a database filled with datasets for finetuning the model
- Creating a finetuned model for summarization of care documents
- Creation of a Summarization web service for on demand care documentation
Results
GPT.3.5-turbo was selected as the best AI model for the projects use case. Data generation involved the creation of realistic datasets to train, validate and test the AI model. The fine-tuning process indicated good performance according to different performance metrics. Furthermore, the development of the web service enabled the mangement of future summary requests in the HerzensApp.
