ReqFusion: A Multi-Provider Framework for Automated PEGS Analysis Across Software Domains
This paper introduces ReqFusion, an AI-enhanced multi-provider framework that leverages multiple Large Language Models and a domain-independent PEGS-based prompting strategy to automate the extraction and classification of software requirements, achieving higher accuracy and significantly reducing analysis time compared to manual methods and generic prompting approaches.
Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine you are trying to build a massive, complex house (a software project). Before you lay a single brick, you need a perfect blueprint. In the world of software, this blueprint is called Requirements Engineering. It's the process of figuring out exactly what the building needs to do, how it should look, and what rules it must follow.
The problem? Reading through hundreds of pages of contracts, proposals, and technical manuals to find these rules is like trying to find a needle in a haystack while wearing blindfolds. It's slow, boring, and humans often miss things or get confused.
Enter ReqFusion, a new AI tool designed to be your super-powered, super-organized assistant for reading these blueprints. Here is how it works, explained simply:
1. The "Super-Reader" Team (Multi-Provider AI)
Imagine you need to read a very difficult legal document. You wouldn't just ask one person to read it; you'd hire three different experts:
- Expert A (OpenAI's GPT): Great at understanding technical jargon and structure.
- Expert B (Anthropic's Claude): Excellent at reading between the lines and understanding subtle context.
- Expert C (Groq's Llama): Fast and efficient for the simpler parts.
ReqFusion doesn't rely on just one "brain." It sends the document to all three AI models at the same time. Then, it acts like a wise judge, comparing their answers.
- If all three agree on a rule, the system is 100% confident it's correct.
- If they disagree, the system flags it for a human to double-check.
The Analogy: Think of it like a panel of judges on a talent show. If all three judges give a "Yes," you know the act is great. If they are split, you know to look closer. This stops the AI from "hallucinating" (making things up), which is a common problem when using just one AI.
2. The "Organized Filing System" (PEGS Method)
Most AI tools are told: "Read this document and tell me what the requirements are." This is like telling a librarian, "Find me some books," without telling them what genre you want. The AI might get overwhelmed or give you a mix of fiction, cookbooks, and history.
ReqFusion uses a specific filing system called PEGS (Project, Environment, Goals, System). Instead of a messy pile, it sorts the information into four neat boxes:
- Project (P): Who is paying? What is the budget? Who are the bosses?
- Environment (E): What outside rules apply? (e.g., "Must work in rain," or "Must follow safety laws").
- Goals (G): What is the big dream? (e.g., "Save 50% of energy").
- System (S): What exactly does the machine do? (e.g., "Press button to start").
The Analogy: Imagine you are packing for a trip. A generic instruction says, "Pack your bag." You might forget your toothbrush or pack a tuxedo for the beach. The PEGS method is like a checklist: "Put toiletries in the blue bag, clothes in the red bag, and documents in the clear folder." Because the AI knows exactly where to look, it finds the right items much faster and more accurately.
3. The Results: Speed, Savings, and Safety
The researchers tested ReqFusion on real-world documents from universities and big German companies. Here is what happened:
- Speed: It did the work 78% faster than a team of human experts. What took humans hours took the AI minutes.
- Accuracy: It found more requirements and made fewer mistakes than a single AI model.
- Cost: It saved about 47% in costs because it used cheaper AI models for simple tasks and saved the expensive ones for the hard work.
Why Does This Matter?
In the past, if a company missed a requirement (like forgetting to mention a safety rule), they might build a software product that fails or is dangerous. ReqFusion acts as a safety net. It ensures that:
- Nothing is missed: It checks every corner of the document.
- Nothing is made up: It cross-checks facts between different AI models.
- Humans stay in charge: If the AI isn't sure, it asks a human for help, rather than guessing.
The Bottom Line
ReqFusion is like giving a construction crew a team of three expert architects who instantly sort the blueprint into perfect categories, double-check each other's work, and hand over a clean, error-free list of instructions. It doesn't replace the human builders; it just makes sure they have the perfect map before they start digging.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.