The VT legacy system, comprising approximately 2.5 million lines of PL/SQL code, lacks consistent documentation and automated tests, posing significant challenges for refactoring and modernisation. This study investigates the feasibility of leveraging large language models (LLMs) to assist in translating PL/SQL code into Java for the modernised "VTF3" system. By leveraging a dataset comprising 10 PL/SQL-to-Java code pairs and 15 Java classes, which collectively established a domain model for the translated files, multiple LLMs were evaluated. Furthermore, we propose a customized prompting strategy that integrates chain-of-guidance reasoning with n-shot prompting. Our findings indicate that this methodology effectively guides LLMs in generating syntactically accurate translations while also achieving functional correctness. However, the findings are limited by the small sample size of available code files and the restricted access to test cases used for validating the correctness of the generated code. Nevertheless, these findings lay the groundwork for scalable, automated solutions in modernising large legacy systems.

The evolution of Python requires accurate version identification to facilitate compatibility and ongoing support. We extend previous work on deep learning models for Python version identification, where LSTM and CodeBERT achieved a 92% accuracy on short code snippets. We further expand these results to larger realistic files, utilising code segmentation techniques for varying input granularities. These techniques ranged from per-line analysis to larger code segments. Our findings show that while LSTM with CodeBERT embeddings maintained high accuracy on short snippets, performance significantly drops on longer segments, particularly in balancing information retention and misclassification risks. Notably, import-statement analysis, despite being the most intuitive indicator of version requirements, reached only a 30% accuracy. This exposes the limitations of our approach when encountering rare or user-defined modules. The findings expose the limitations of deep learning for language version identification, and suggest that alternative approaches may be necessary for high accuracy on larger datasets.