Software Engineering Assignment: Development of an AI-Powered Requirement Elicitation Agent for SMEs

• Classification Accuracy: For FR, NFR, and Ambiguity classification.

3. Experimental Setup and Execution

a. AI Model Selection:

• Use transformer-based models (e.g., BERT, RoBERTa) fine-tuned for:
• Multi-class classification (FR/NFR/Ambiguous)
• Text generation (for SRS drafting and question generation)

b. Tools and Frameworks:

• NLP Frameworks: HuggingFace Transformers, spaCy, NLTK

4. Innovation and Research Contribution

a. Novelty:

• Combines requirement classification, ambiguity detection, and real-time stakeholder interaction in a single AI agent.
• Introduces an automated clarification mechanism, which is often manual in traditional requirement elicitation.

b. Contribution:

• A prototype that can generate a structured SRS from informal inputs via chat or voice.
• Possibly a new labeled dataset for FR/NFR/ambiguity classification, which can be published for future research.

Tools:

GPT-4 API or open-source LLM (LLaMA, Mistral)
LangChain for multi-turn dialogue
Streamlit or React + Flask for front-end

CA1: diverse resources (people, money, equipment, materials, information, and technologies) Project involves of diverse resources such as software, hardware, data, research materials, guidance from the experts in the field.

CA3: involves creativity and innovation in providing a solution The implemented solution meets the needs of the target audience or user. This involves gathering feedback and input from users throughout the design and development process. Students need to identify innovative solutions to the problem they are trying to solve by incorporating new technologies, engaging user in the interaction with the developed solution, or create new method/ approach. Combining knowledge and skills from different fields can create a more innovative and effective solution.