# Requirements Engineering using Generative AI: Prompts and Prompting Patterns

**Authors:** Krishna Ronanki, Beatriz Cabrero-Daniel, Jennifer Horkoff, Christian Berger  
**arXiv:** [2311.03832](https://arxiv.org/abs/2311.03832)  
**Date:** 2023-11-07  
**Summary page:** https://www.summarizepaper.com/en/paper/2311.03832v1/

## Key points

- Study focused on automating Requirements Engineering (RE) tasks using GenAI
- Utilized GPT-3.5 turbo API for automating requirement classification and tracing tasks
- Employed five different prompt patterns for effective task performance
- Dataset of System Requirements Specification (SRS) documents used for ground truth in requirement classification
- Binary Classification task aimed to distinguish Functional Requirements (FR) from Non-Functional Requirements (NFR)
- Requirements Traceability task involved identifying related requirements in SRS documents
- Performance evaluation metrics included precision, recall, accuracy, and F-Score to assess prompt patterns effectiveness
- Recommendations provided on selecting appropriate prompt patterns for specific RE tasks
- Evaluation framework offered for researchers and practitioners in the field
- Manual formatting of SRS files and removal of unnecessary information conducted for consistency in evaluation process
- Information Retrieval measures like precision, recall, and F-measure used to evaluate tool effectiveness in RE field

## Summary

The study focused on automating Requirements Engineering (RE) tasks using GenAI and evaluating prompt patterns for effective task performance. The researchers utilized the GPT-3.5 turbo API to automate requirement classification and tracing tasks, employing five different prompt patterns. For requirement classification, a dataset of System Requirements Specification (SRS) documents with trace links was used to establish ground truth. The Binary Classification task aimed to distinguish Functional Requirements (FR) from Non-Functional Requirements (NFR), while the Requirements Traceability task involved identifying related requirements in SRS documents. Performance evaluation metrics such as precision, recall, accuracy, and F-Score were used to assess the effectiveness of prompt patterns in executing the RE tasks. The study provided recommendations on selecting appropriate prompt patterns for specific RE tasks and offered an evaluation framework for researchers and practitioners. Additionally, manual formatting of SRS files and removal of unnecessary information were conducted to ensure consistency in the evaluation process. The study adopted Information Retrieval measures like precision, recall, and F-measure to evaluate tool effectiveness in the RE field. Overall, the research contributes valuable insights into automating RE tasks using GenAI and highlights the importance of prompt engineering for successful automation in Requirements Engineering.