Timeline of the thesis:
- We tested the ability of Llama 2 on 0-shot text generation for our task. For comparison we also tested with another LLM - T5 (question_generation_no_finetune)
- Finetuned Llama 2 on the dataset and try again (question_generation_finetune)
- Test + finetune Llama 2 on trigger detection task (triggers_detection_finetune)
- Prepared the auxiliary database (auxiliary_data_prep)
- Processed the auxiliary data and the ehr (ehr_processing_and_final_question_generation)