semantica/docs/reference/kg.md

Knowledge Graph

High-level KG construction, management, and analysis system.

---

🎯 Overview

• :material-graph-outline:{ .lg .middle } KG Construction

  ---

  Build graphs from entities and relationships with automatic merging

• :material-clock-time-four-outline:{ .lg .middle } Temporal Graphs

  ---

  Time-aware edges (valid_from, valid_until) and temporal queries

• :material-account-multiple-check:{ .lg .middle } Entity Resolution

  ---

  Deduplicate entities using fuzzy matching and semantic similarity

• :material-alert-decagram:{ .lg .middle } Conflict Detection

  ---

  Detect and resolve contradicting facts from multiple sources

• :material-chart-network:{ .lg .middle } Graph Analytics

  ---

  Centrality, Community Detection, and Connectivity analysis

• :material-history:{ .lg .middle } Provenance

  ---

  Track the source and lineage of every node and edge

!!! tip "When to Use" - KG Building: The primary module for assembling a KG from extracted data - Data Cleaning: Merging duplicates and resolving conflicts - Analysis: Understanding the structure and importance of nodes - Time-Series: Modeling how the graph evolves over time

---

⚙️ Algorithms Used

Entity Resolution

• Fuzzy Matching: Levenshtein/Jaro-Winkler distance for string similarity.
• Semantic Matching: Cosine similarity of embeddings.
• Transitive Merging: If A=B and B=C, then A=B=C.

Graph Analytics

• Centrality: Degree, Betweenness, Closeness, Eigenvector.
• Communities: Louvain, Leiden, K-Clique.
• Connectivity: Connected Components, Bridge Detection.

Temporal Analysis

• Time-Slicing: Viewing the graph at a specific point in time.
• Interval Algebra: Allen's interval algebra for temporal reasoning (overlaps, during, before).

---

Main Classes

GraphBuilder

Constructs the KG from raw data.

Methods:

Method	Description
build(sources)	Build graph from inputs
merge_entities()	Run deduplication

Example:

from semantica.kg import GraphBuilder

builder = GraphBuilder(merge_entities=True)
kg = builder.build([source1, source2])

GraphAnalyzer

Runs analytical algorithms.

Methods:

Method	Description
centrality(method)	Calculate importance
communities(method)	Find clusters

TemporalGraphQuery

Queries time-aware graphs.

Methods:

Method	Description
at_time(timestamp)	Graph state at T
during(start, end)	Graph state in interval

---

Convenience Functions

from semantica.kg import GraphBuilder, analyze_graph

# Build using GraphBuilder
builder = GraphBuilder(resolve_conflicts=True)
kg = builder.build(sources)

# Analyze
stats = analyze_graph(kg)
print(f"Communities: {stats['communities']}")

---

Configuration

Environment Variables

export KG_MERGE_STRATEGY=fuzzy
export KG_TEMPORAL_GRANULARITY=day
export KG_CONFLICT_RESOLUTION=confidence

YAML Configuration

kg:
  resolution:
    threshold: 0.9
    strategy: semantic
    
  temporal:
    enabled: true
    default_validity: infinite

---

Integration Examples

Temporal Analysis Pipeline

from semantica.kg import GraphBuilder, TemporalGraphQuery

# 1. Build Temporal Graph
builder = GraphBuilder(enable_temporal=True)
kg = builder.build(temporal_data)

# 2. Query Evolution
query = TemporalGraphQuery(kg)
snapshot_2020 = query.at_time("2020-01-01")
snapshot_2023 = query.at_time("2023-01-01")

# 3. Compare
diff = snapshot_2023.minus(snapshot_2020)
print(f"New nodes since 2020: {len(diff.nodes)}")

---

Best Practices

1. Clean Data First: Use EntityResolver aggressively to prevent "entity explosion" (too many duplicate nodes).
2. Use Provenance: Always track sources (track_history=True) to debug where bad data came from.
3. Temporal Granularity: Choose the right granularity (Day vs Second) to balance performance and precision.
4. Validate: Run GraphValidator after building to ensure structural integrity.

---

See Also

• Graph Store Module - Persistence layer
• Semantic Extract Module - Data source
• Visualization Module - Visualizing the KG