As a research assistant, I was responsible for the dataset preparation and training process of the model, collecting and evaluating the information from websites and generating a comprehensive dataset for model training use. My biggest contribution was scripting various data augmentation methods used for our dataset to achieve a large enough dataset of the niche topic for better training performance, I used my computing, data science, research, problem-solving skills and technical knowledge of NPL. I was in constant communication with colleagues from various projects to obtain useful insights and with users to receive feedback on areas needing improvement.
I had several individual tasks including creating trial transformer models and generating an annotated question-answering database etc., where I applied many Design Engineering skills like HCD skills to think from the user’s perspective. My responsibilities from these tasks included data collection, annotation and argumentation, model training and validation, chatbot safety and integrity testing. The contribution I made was crucial in boosting the chatbot's precision and raising the level of user engagement.
Other responsibilities included leading communication with our partner NHS staff and Asthma+Lung UK to gain authority and validation for my dataset and writing a conference paper with my supervisor to submit our work on data augmentation.
