analysis/scorer.py

"""
Understanding Scorer Module
Calculates final understanding scores from analysis results
"""

from typing import Dict, Optional

class UnderstandingScorer:
    def __init__(self):
        # Scoring weights - adjusted these a few times based on testing
        # coverage is weighted highest because it's most reliable
        self.weights = {
            'consistency': 0.25,
            'coverage': 0.35,
            'stability': 0.25,
            'assumptions': 0.15
        }
    
    def calculate_scores(
        self,
        consistency_result: Dict,
        coverage_result: Dict,
        stability_result: Optional[Dict]
    ) -> Dict:
        """
        Calculate multi-dimensional understanding scores
        
        Returns:
            {
                'overall': float (0-100),
                'consistency': float (0-100),
                'coverage': float (0-100),
                'stability': float (0-100),
                'assumptions': float (0-100)
            }
        """
        # Extract individual scores
        consistency_score = consistency_result.get('consistency_score', 0)
        coverage_score = coverage_result.get('coverage_score', 0)
        stability_score = stability_result.get('stability_score', 100) if stability_result else 100
        
        # Calculate assumption completeness score
        # Based on whether key assumptions are made explicit
        assumptions_score = self._calculate_assumption_score(
            coverage_result=coverage_result,
            consistency_result=consistency_result
        )
        
        # Calculate weighted overall score
        overall_score = (
            consistency_score * self.weights['consistency'] +
            coverage_score * self.weights['coverage'] +
            stability_score * self.weights['stability'] +
            assumptions_score * self.weights['assumptions']
        )
        
        return {
            'overall': round(overall_score, 1),
            'consistency': round(consistency_score, 1),
            'coverage': round(coverage_score, 1),
            'stability': round(stability_score, 1),
            'assumptions': round(assumptions_score, 1)
        }
    
    def _calculate_assumption_score(
        self,
        coverage_result: Dict,
        consistency_result: Dict
    ) -> float:
        """
        Calculate assumption completeness score
        
        High score = explicit about assumptions, boundary conditions, limitations
        Low score = makes implicit assumptions without stating them
        """
        # Simple heuristic for now - basically if consistency and coverage are good,
        # assumptions are probably okay. Not perfect but works decently.
        
        consistency_score = consistency_result.get('consistency_score', 0)
        coverage_score = coverage_result.get('coverage_score', 0)
        
        # Missing concepts indicate unstated assumptions
        missing_count = len(coverage_result.get('missing_concepts', []))
        
        # Contradictions indicate conflicting implicit assumptions
        contradiction_count = len(consistency_result.get('contradictions', []))
        
        # Base score on coverage and consistency
        base_score = (consistency_score + coverage_score) / 2
        
        # Penalize for missing concepts (unstated prerequisites)
        missing_penalty = min(30, missing_count * 10)
        
        # Penalize for contradictions (conflicting assumptions)
        contradiction_penalty = min(20, contradiction_count * 15)
        
        final_score = base_score - missing_penalty - contradiction_penalty
        
        return max(0, min(100, final_score))