Automatic source code summarization is the task of generating a readable summary that describes the functionality of the code in natural language. In recent years, the use of deep learning-based approaches has led to significant improvement in the performance of automatic code summarization, e.g., using Transformers and Graph Neural Networks. However, the performance is still far from optimal and developers that are unsatisfied with a given summary are not able to provide feedback or additional information that can be used to refine the output.

In this research project, the goal is to investigate ways in which additional input from the developer can further improve the performance of automatic code summarization. Specifically, the main tasks in the project are:

1. Investigating existing failures of state-of-the-art source code summarization solutions
2. Developing new computational approaches and interactive schemes for incorporating developer input and feedback in order to improve the performance of deep learning-based approaches for source code summarization
3. Evaluating the impact of the new approaches on existing large code summarization datasets.

The responsibilities of the SUDS student will be:

1. Read about, implement, and empirically evaluate state-of-the-art models for automatic source code summarization.
2. Investigate existing failures of state-of-the-art source code summarization solutions and develop interactive schemes for incorporating developer input and feedback in order to improve their performance.

Evaluating the impact of the new approaches on existing large code summarization datasets.