Evidence Classification

Evidence Classification

Evidence Classification is the system layer that assigns structured categories to ingested evidence, ensuring each unit is properly typed before scoring, validation, and ranking processes.

Context Block

Page Type: Evidence System Layer
Function: Classification Engine
Position: After Evidence Ingestion
Role: Converts raw evidence objects into structured categories

This layer ensures that all evidence is not treated uniformly. Instead, it is segmented based on epistemic function and structural characteristics.

Core Objective

  • Assign categorical labels to evidence units
  • Separate factual, analytical, and synthetic evidence
  • Prepare evidence for scoring and ranking layers
  • Improve retrieval-to-evidence mapping precision
  • Reduce epistemic ambiguity in downstream systems

Classification Pipeline

1. Feature Extraction
Extracts structural and semantic features from ingested evidence.

2. Type Detection
Determines evidence nature based on source and content structure.

3. Epistemic Role Assignment
Defines role of evidence in reasoning (supporting, primary, contextual).

4. Category Mapping
Maps evidence to predefined classification schema.

5. Output Structuring
Produces standardized classified evidence objects.

Evidence Types

  • Primary Evidence — direct facts, raw data, official sources
  • Secondary Evidence — interpretation, analysis, summaries
  • Tertiary Evidence — aggregated insights, compiled knowledge
  • Synthetic Evidence — model-generated or inferred signals

Epistemic Roles

  • Core Evidence — primary grounding source
  • Supporting Evidence — reinforces core claims
  • Contextual Evidence — provides background understanding
  • Contradictory Evidence — introduces alternative truth signals

Classification Signals

  • Source authority level
  • Data structure type (structured/unstructured)
  • Intent alignment strength
  • Ontology mapping compatibility
  • Information density score

Integration in GEO Pipeline

Evidence Classification acts as the structural bridge between raw ingestion and quantitative evaluation layers.

Failure Modes

  • Misclassification of synthetic vs factual evidence
  • Over-generalization of evidence categories
  • Loss of epistemic role distinctions
  • Incorrect mapping to ontology-compatible types

Structured Output Model

Each evidence unit produces:

  • Evidence Type Label
  • Epistemic Role Assignment
  • Category Confidence Score
  • Structural Feature Map
  • Classification Trace ID

Relationship Block

Parent Layer: /evidence/
Upstream: Evidence Ingestion
Downstream: Evidence Scoring, Evidence Ranking
Connected Systems: Ontology Layer, Retrieval Engine, Answer System

Structured Summary

Evidence Classification is the structural categorization layer of the Evidence system. It transforms raw evidence into epistemically meaningful types that define how each piece of information should be evaluated and used.

This layer ensures downstream systems operate on well-structured and semantically distinct evidence units.