Towards a Conversational User Interface for Aiding Researchers with Reproducibility
Lázaro Costa · 2024
In science, it is very important to be able to recreate the same computing environment to reproduce the same results achieved by previous scientific experiments. However, it is challenging to create a shareable package with the same environment using the same programming languages, frameworks, or data sources due to the diversity of researchers’ knowledge and the variety of computational environments used. In this work, I propose designing and constructing a conversational user interface that allows researchers to upload experiment files and clarify the necessary information via text communication to create a reproducible experiment package. My approach uses an integrated Large Language Model (LLM) that allows the platform to infer, whenever possible, some information (e.g., the programming language used, the main file to be executed, the parameters needed to be inserted when executing) to reduce the amount of information asked to the user. With this work, I intend to use an LLM to guide the researcher through this procedure, thereby reducing the time spent and the number of interactions required to create a reproducible package.