Inclusive Requirements Elicitation
AI in Libraries to Support Respectful Management of Indigenous Knowledges
Samantha Callaghan, Arianna Ciula, Tiffany Ong, Miguel Vieira
Context for Requirements Elicitation
- iREAL: Indigenous data considerations
- User research (to be re-integrated)
- Workshop dashboard constraints
- Today's focus:
- Identify use cases and requirements
Requirements
- What is a requirement?
- Purpose: How does it fit project/user needs?
- Definition: How do we specify them?
- Prioritisation: How do we rank importance?
MoSCoW Prioritisation
- Must have
- Should have
- Could have
- Won't have this time
- Defines out-of-scope items
- Potential for future phases
MoSCoW Prioritisation
System Architecture
- School records: NSW Aboriginal school records (1876-1979)
- Data preparation: Conversion to JSON and Markdown
- Data extraction: AI/ML models for metadata extraction
- Dashboard: Frontend to explore processed data
Data Preparation
-
Original documents converted to:
- JSON (for data extraction)
- Markdown (for frontend rendering)
-
Each record divided into smaller, manageable sections
-
Each section is processed for data extraction
- NER: SpanMarker model
- Geocoding: Custom transformer
- Keywords: LLM (Mistral 7B) with custom prompt for 5 keywords per text chunk
- Topics: LLM with custom prompt for overarching themes
