Towards Automated Knowledge Management in the Software Life Cycle

Artificial Intelligence (AI), in particular that built upon Large Language Models (LLMs), is reshaping the way we approach many processes. Mainly thanks to the introduction of coding assistants, software development has been entangled in this transformation as well. However, when we talk about software development, coding is just a part of this intricate process, and it is not the only one that can benefit from the support of AI. In fact, LLMs are already being successfully applied to other stages of the software life cycle, such as requirement analysis and design. With this paper, we focus on the big picture of the entire life cycle of software and on maintaining structured knowledge about the system, with attention on maintaining the consistency with respect to formal and informal specifications across all artefacts. The core of our idea is to involve LLM-based assistants within all the life cycle stages, where information to keep track of can come from both well-structured artefact sources (e.g., in source code and documentation) as well as sparse and unstructured ones (e.g., in meeting transcripts). In fact, these assistants can potentially be used to automate information extraction from multiple different sources and to maintain a Knowledge Base (KB) modelling the knowledge about the system that can be used to ensure continuous consistency of the project artefacts through validation against the KB itself. ... mehrThe role we envision for AI is thus (i) to bridge the gap between structured and unstructured knowledge and its formal representation and (ii) to automate the update of information and the verification of consistency.