This resource is related to the research paper "Towards Code Summarization for Scientific Domain Experts on Scarce Data" created as part of the PhD thesis*. 

Nowadays, programming has spread into nearly all expert domains. Programming code is no longer written only by software engineers and developers but also by domain experts, who talk about the code and search for it differently than IT specialists. This results in the need for domain expert coding search systems that address their custom way of working with code. Frequently, these systems rely on automatic source code summarization, which needs to be adjusted to the domain knowledge.

This work proposes the Algorithm Choice Framework for code summarization in the science domain, which aims to support domain experts in navigating the surface of algorithm selection, prioritizing the low extent of programming to achieve the task. Based on the code summarization research and MLOps workflows, this framework identifies the criteria influencing the model choice in the current period and beyond, making the algorithms’ choice robust towards the potential changes in the requirements and environment. The framework usage was finally exemplified in a real-world use case – Quantlet.com, the scientific coding university research database and search mechanism focusing on quantitative research and finance.

*Work is not related to Amazon.

The code is available in Quantlet org at Github and is searchable through Quantlet