Search for: All records

Software documentation supports a broad set of software maintenance tasks; however, creating and maintaining high-quality, multi-level software documentation can be incredibly time-consuming and therefore many code bases suffer from a lack of adequate documentation. We address this problem through presenting HGEN, a fully automated pipeline that leverages LLMs to transform source code through a series of six stages into a well-organized hierarchy of formatted documents. We evaluate HGEN both quantitatively and qualitatively. First, we use it to generate documentation for three diverse projects, and engage key developers in comparing the quality of the generated documentation against their own previously produced manually-crafted documentation. We then pilot HGEN in nine different industrial projects using diverse datasets provided by each project. We collect feedback from project stakeholders, and analyze it using an inductive approach to identify recurring themes. Results show that HGEN produces artifact hierarchies similar in quality to manually constructed documentation, with much higher coverage of the core concepts than the baseline approach. Stakeholder feedback highlights HGEN's commercial impact potential as a tool for accelerating code comprehension and maintenance tasks. Results and associated supplemental materials can be found at https://zenodo.org/records/11403244. 

Software engineering practices such as constructing requirements and establishing traceability help ensure systems are safe, reliable, and maintainable. However, they can be resource-intensive and are frequently underutilized. To alleviate the burden of these essential processes, we developed the Requirements Organization and Optimization Tool (ROOT). ROOT centralizes project information and offers project visualizations and AI-based tools designed to streamline engineering processes. With ROOT's assistance, engineers benefit from improved oversight and early error detection, leading to the successful development of software systems. A link to a screen cast can be found at: https://youtu.be/3rtMYRnsu24 

Instruction tuning has remarkably advanced large language models (LLMs) in understand- ing and responding to diverse human instruc- tions. Despite the success in high-resource lan- guages, its application in lower-resource ones faces challenges due to the imbalanced foun- dational abilities of LLMs across different lan- guages, stemming from the uneven language distribution in their pre-training data. To tackle this issue, we propose pivot language guided generation (PLUG), an approach that utilizes a high-resource language, primarily English, as the pivot to enhance instruction tuning in lower-resource languages. It trains the model to first process instructions in the pivot language, and then produce responses in the target lan- guage. To evaluate our approach, we introduce a benchmark, X-AlpacaEval, of instructions in 4 languages (Chinese, Korean, Italian, and Spanish), each annotated by professional trans- lators. Our approach demonstrates a significant improvement in the instruction-following abili- ties of LLMs by 29% on average, compared to directly responding in the target language alone. Further experiments validate the versatility of our approach by employing alternative pivot languages beyond English to assist languages where LLMs exhibit lower proficiency. 

Automatic taxonomy induction is crucial for web search, recommendation systems, and question answering. Manual curation of taxonomies is expensive in terms of human effort, making automatic taxonomy construction highly desirable. In this work, we introduce Chain-of-Layer which is an in-context learning framework designed to induct taxonomies from a given set of entities. Chain-of-Layer breaks down the task into selecting relevant candidate entities in each layer and gradually building the taxonomy from top to bottom. To minimize errors, we introduce the Ensemble-based Ranking Filter to reduce the hallucinated content generated at each iteration. Through extensive experiments, we demonstrate that Chain-of-Layer achieves state-of-the-art performance on four real-world benchmarks. 

Large Language Models (LLMs) have the potential to revolutionize automated traceability by overcoming the challenges faced by previous methods and introducing new possibilities. However, the optimal utilization of LLMs for automated traceability remains unclear. This paper explores the process of prompt engineering to extract link predictions from an LLM. We provide detailed insights into our approach for constructing effective prompts, offering our lessons learned. Additionally, we propose multiple strategies for leveraging LLMs to generate traceability links, improving upon previous zero-shot methods on the ranking of candidate links after prompt refinement. The primary objective of this paper is to inspire and assist future researchers and engineers by highlighting the process of constructing traceability prompts to effectively harness LLMs for advancing automatic traceability. 

Taxonomies serve many applications with a structural representation of knowledge. To incorporate emerging concepts into existing taxonomies, the task of taxonomy completion aims to find suitable positions for emerging query concepts. Previous work captured homogeneous token-level interactions inside a concatenation of the query concept term and definition using pre-trained language mod- els. However, they ignored the token-level interactions between the term and definition of the query concepts and their related concepts. In this work, we propose to capture heterogeneous token-level interactions between the different textual components of concepts that have different types of relations. We design a relation-aware mutual attention module (RAMA) to learn such interactions for taxonomy completion. Experimental results demonstrate that our new taxonomy completion framework based on RAMA achieves the state-of-the-art performance on six taxonomy datasets.