TKG (Temporal Knowledge Graph)

Overview

TKG (Temporal Knowledge Graph) is a knowledge graph that treats time and causality as first-class information. Unlike a regular Knowledge Graph, it stores not only how facts evolve over time (when they are true) but also the causal relationships between facts (what caused what) — so AI can answer temporal and causal questions accurately.

Why Temporal Information Matters

Facts change over time. For example:

• A company's CEO can change quarterly.
• Exchange rates fluctuate every second.
• Laws and regulations get amended.
• Research findings get revised with new discoveries.

How TKG Handles Time

TKG attaches both an event time and a validity interval to each fact.

• Event Time — When the event the fact refers to occurred
• Valid From — When the fact became valid
• Valid Until — When the fact stopped being valid (none means still valid)

Reads always happen as of a specific point in time: only facts that were valid at that moment are returned, with no information that became known later. This makes it possible to reproduce any past moment exactly, free of hindsight bias.

Causal Relationships

TKG links facts with typed causal edges — not a vague "related to" but an explicit kind of relationship.

• Causes — One fact directly brings about another
• Triggers — Sets an event in motion
• Contributes to — Partially contributes to an outcome
• Prevents — Keeps something from happening

Each causal link carries its mechanism, the time lag between cause and effect, and a confidence score. Newly discovered links enter as hypotheses and are promoted to verified only after testing. This lets TKG separate mere correlation from genuine causation and give grounded answers to "why" and "what caused this" questions.

Benefits

• Accurate Verification — Verify facts as of a specific point in time.
• Change History Tracking — See the complete history of how facts have changed.
• Conflict Detection — Automatically detect conflicts between facts from different times.
• Causal Explanation — Trace not just outcomes but their causes and mechanisms to answer "why."