Justin F. Brunelle

Colin Diggs, Michael Doyle, Amit Madan et al.

Legacy software systems, written in outdated languages like MUMPS and mainframe assembly, pose challenges in efficiency, maintenance, staffing, and security. While LLMs offer promise for modernizing these systems, their ability to understand legacy languages is largely unknown. This paper investigates the utilization of LLMs to generate documentation for legacy code using two datasets: an electronic health records (EHR) system in MUMPS and open-source applications in IBM mainframe Assembly Language Code (ALC). We propose a prompting strategy for generating line-wise code comments and a rubric to evaluate their completeness, readability, usefulness, and hallucination. Our study assesses the correlation between human evaluations and automated metrics, such as code complexity and reference-based metrics. We find that LLM-generated comments for MUMPS and ALC are generally hallucination-free, complete, readable, and useful compared to ground-truth comments, though ALC poses challenges. However, no automated metrics strongly correlate with comment quality to predict or measure LLM performance. Our findings highlight the limitations of current automated measures and the need for better evaluation metrics for LLM-generated documentation in legacy systems.

Christopher Glasz, Emily Escamilla, Eric O. Scott et al.

Large language models (LLMs) have become essential tools in computer science, especially for tasks involving code understanding and generation. However, existing work does not address many of the unique challenges presented by code written for government applications. In particular, government enterprise software is often written in legacy languages like MUMPS or assembly language code (ALC) and the overall token lengths of these systems exceed the context window size for current commercially available LLMs. Additionally, LLMs are primarily trained on modern software languages and have undergone limited testing with legacy languages, making their ability to understand legacy languages unknown and, hence, an area for empirical study. This paper examines the application of LLMs in the modernization of legacy government code written in ALC and MUMPS, addressing the challenges of input limitations. We investigate various code-chunking methods to optimize the generation of summary module comments for legacy code files, evaluating the impact of code-chunking methods on the quality of documentation produced by different LLMs, including GPT-4o, Claude 3 Sonnet, Mixtral, and Llama 3. Our results indicate that LLMs can select partition points closely aligned with human expert partitioning. We also find that chunking approaches have significant impact on downstream tasks such as documentation generation. LLM-created partitions produce comments that are up to 20% more factual and up to 10% more useful than when humans create partitions. Therefore, we conclude that LLMs can be used as suitable replacements for human partitioning of large codebases during LLM-aided modernization.

Rock Sabetto, Emily Escamilla, Devesh Agarwal et al.

Large language models (LLMs) have been increasingly integrated into software engineering and maintenance tasks due to their high performance with software engineering tasks and robust understanding of modern programming languages. However, the ability of LLMs to comprehend code written with legacy languages remains a research gap challenged by real-world legacy systems lacking or containing inaccurate documentation that may impact LLM comprehension. To assess LLM comprehension of legacy languages, there is a need for objective LLM evaluation. In order to objectively measure LLM comprehension of legacy languages, we need an efficient, quantitative evaluation method. We leverage multiple-choice question answering (MCQA), an emerging LLM evaluation methodology, to evaluate LLM comprehension of legacy code and the impact of comment prevalence and inaccurate comments. In this work, we present preliminary findings on the impact of documentation on LLM comprehension of legacy code and outline strategic objectives for future work.

Trevor Bostic, Jeff Stanley, John Higgins et al.

The web is the prominent way information is exchanged in the 21st century. However, ensuring web-based information is accessible is complicated, particularly with web applications that rely on JavaScript and other technologies to deliver and build representations; representations are often the HTML, images, or other code a server delivers for a web resource. Static representations are becoming rarer and assessing the accessibility of web-based information to ensure it is available to all users is increasingly difficult given the dynamic nature of representations. In this work, we survey three ongoing research threads that can inform web accessibility solutions: assessing web accessibility, modeling web user activity, and web application crawling. Current web accessibility research is continually focused on increasing the percentage of automatically testable standards, but still relies heavily upon manual testing for complex interactive applications. Along-side web accessibility research, there are mechanisms developed by researchers that replicate user interactions with web pages based on usage patterns. Crawling web applications is a broad research domain; exposing content in web applications is difficult because of incompatibilities in web crawlers and the technologies used to create the applications. We describe research on crawling the deep web by exercising user forms. We close with a thought exercise regarding the convergence of these three threads and the future of automated, web-based accessibility evaluation and assurance through a use case in web archiving. These research efforts provide insight into how users interact with websites, how to automate and simulate user interactions, how to record the results of user interactions, and how to analyze, evaluate, and map resulting website content to determine its relative accessibility.

Justin F. Brunelle, Michele C. Weigle, Michael L. Nelson

Web resources are increasingly interactive, resulting in resources that are increasingly difficult to archive. The archival difficulty is based on the use of client-side technologies (e.g., JavaScript) to change the client-side state of a representation after it has initially loaded. We refer to these representations as deferred representations. We can better archive deferred representations using tools like headless browsing clients. We use 10,000 seed Universal Resource Identifiers (URIs) to explore the impact of including PhantomJS -- a headless browsing tool -- into the crawling process by comparing the performance of wget (the baseline), PhantomJS, and Heritrix. Heritrix crawled 2.065 URIs per second, 12.15 times faster than PhantomJS and 2.4 times faster than wget. However, PhantomJS discovered 531,484 URIs, 1.75 times more than Heritrix and 4.11 times more than wget. To take advantage of the performance benefits of Heritrix and the URI discovery of PhantomJS, we recommend a tiered crawling strategy in which a classifier predicts whether a representation will be deferred or not, and only resources with deferred representations are crawled with PhantomJS while resources without deferred representations are crawled with Heritrix. We show that this approach is 5.2 times faster than using only PhantomJS and creates a frontier (set of URIs to be crawled) 1.8 times larger than using only Heritrix.