Question Generation System

A comprehensive question generation system for state government assistant evaluation, supporting multiple states with extensive template libraries.

Features

• Multi-State Support: SC, HI, IN, MS with state-specific question templates
• 100+ Unique Questions: Expanded template library supporting large question sets
• Category Coverage: 10+ categories including government, business, employment, healthcare, education, transportation, housing, taxation, recreation, and seniors
• Complexity Levels: Basic, intermediate, and complex questions with configurable distributions
• User Personas: 9 different user personas (general, senior_citizen, small_business_owner, etc.)
• Question Variations: Automatic generation of question variations for diversity

Files

Core Generator

• simple_question_generator.py: Main question generation engine without expected links logic
• generate_simple_questions.py: Command-line interface wrapper

Usage

Basic Usage

# Generate 25 questions for South Carolina
python3 generate_simple_questions.py --count 25 --state SC --state-name "South Carolina"

# Generate 100 questions for Mississippi  
python3 generate_simple_questions.py --count 100 --state MS --state-name Mississippi

# Generate focused question sets
python3 generate_simple_questions.py --count 50 --question-type basic_services --state HI --state-name Hawaii

Parameters

• --count: Number of questions to generate (default: 25)
• --question-type: Type of questions (comprehensive, basic_services, complex_scenarios)
• --state: State code (SC, HI, IN, MS)
• --state-name: Full state name for question customization

Supported States

• SC (South Carolina): Full template library with 350+ base questions
• HI (Hawaii): Includes environment category for DLNR focus
• IN (Indiana): Standard categories with state-specific templates
• MS (Mississippi): Standard categories with state-specific templates

Question Categories

1. Government: Voting, licenses, permits, records, elected officials
2. Business: Registration, licenses, permits, taxes, regulations
3. Employment: Unemployment, job search, workers comp, labor rights
4. Healthcare: Medicaid, insurance, public health, mental health
5. Education: K12, higher ed, financial aid, adult education
6. Transportation: DMV, public transit, roads, vehicle registration
7. Housing: Assistance, regulations, property tax, first-time buyers
8. Taxation: Income tax, property tax, business tax, assistance
9. Recreation: Parks, hunting/fishing, tourism, events
10. Seniors: Benefits, healthcare, housing, transportation
11. Environment (Hawaii only): Conservation, marine life, permits

Output Format

Generated questions are saved as JSON files with the following structure:

{
  "id": "Q001",
  "question": "How do I register to vote in South Carolina?",
  "category": "government",
  "subcategory": "voting",
  "complexity": "basic",
  "priority": "high", 
  "user_persona": "new_resident",
  "state": "SC"
}

Template Expansion

The system includes an expanded template library with:

• 15+ questions per complexity level for each category
• 350+ total base templates across all categories
• Question variations that modify phrasing for diversity
• State-specific customization for relevant agencies and processes

Examples

Generate Comprehensive Question Set

python3 generate_simple_questions.py --count 50 --state SC --state-name "South Carolina"

Generate Basic Services Focus

python3 generate_simple_questions.py --count 30 --question-type basic_services --state IN --state-name Indiana

Generate Complex Scenarios

python3 generate_simple_questions.py --count 25 --question-type complex_scenarios --state MS --state-name Mississippi

Development

The question generator uses template-based generation with:

• Random selection from category-specific templates
• Complexity distribution based on configurable weights
• Question variations for natural language diversity
• State-specific templates for accurate local context

No external APIs or AI models required - fully deterministic generation based on predefined templates.