← Back to concept explainers

Concept Fable

Triple

Conclusion first:A triple is usually expressed in a "subject-predicate-object" structure to represent a fact or relationship, for example "Shopkeeper Zhao-operates-Linji Tea House."

Triple · Knowledge representation

Read the story first. There is no rush to give a definition here; let the question grow on its own first.

Fable

Three seals are required for one fact

The county archive was piled full of contracts and rumors. When someone asked, "To whom was the Liu family tea garden sold?", the clerk had to search for half a day, because a single sentence was often buried in a long document.

The new curator had the scribes break facts into three seals: who, what relationship, and who or what object. For example, "Liu family tea garden—sold to—Qian family," "Manager Zhao—lives on—East Street." Each fact was like a small bamboo tag that could be inserted into the relationship network.

At first the scribe cared only about speed and stamped rumors as facts too. Errors soon appeared: someone said the tea garden was leased to the Qian family, but the register wrote it as sold to the Qian family; another entry came from a ten-year-old contract and had already expired.

The curator added a marginal note: beside each group of three seals, the source, time, credibility level, and visibility scope must be written. Those from official contracts could be made public, those from private letters had to be reviewed, and those involving privacy could only be viewed internally.

A few months later, the archive room could answer more complex questions: which tea gardens the Qian family had bought, which shops on East Street belonged to the same person, which relationships had lapsed last year. Answers no longer relied on full-text luck, but on clear relationships linked one by one.

At first the scribe found it troublesome and only copied whole sentences onto slips of paper. When checking "who bought the tea garden," the slips still had to be read one by one; when checking "what else did the Qian family buy," it became an even bigger mess. The text had been moved, but the relationships had not been extracted.

The curator required that each bamboo slip record only one clear relationship and be able to connect with other slips. The Liu family tea garden sold to the Qian family, the Qian family was on South Street, and South Street was under East Ward. That way, a question could travel outward along the slips and find a network of people, things and relationships. The archive no longer stored only sentences, but connectable facts. Later, when newcomers joined the archive, they first practiced breaking down ten facts, then practiced linking the slips back to the original text. Breaking down too coarsely severs meaning; breaking down clearly makes inquiry easy.

The scribe understood that archives are not just collections of text. Only by breaking the world into small facts of "subject, relation, object" and marking each fact with its source and boundary can the relationship network be queryable, updatable and accountable. The clearer the three seals, the more stable the later reasoning.

Reveal

This story is about: triples

A triple is usually expressed in a "subject-predicate-object" structure to represent a fact or relationship, for example "Shopkeeper Zhao-operates-Linji Tea House."

Triples are commonly used in knowledge graphs, relation extraction, and semantic data. A truly usable triple also needs source, time, confidence, version, and permissions; otherwise the system may treat expired, rumored, or non-public information as established fact.

Metaphor mapping

  • The first seal: the subject, the party that initiates the relationship
  • Second seal: predicate, describing the relationship type
  • The third seal: object, the target to which the relationship points
  • Source, time, confidence: fact metadata
  • Public or internal viewing: the permission boundary of triples

Andy Ren Xu's judgment

The value of triples is turning facts in text into a structure the system can query, connect and reason over. Enterprise knowledge bases are not short on documents; what they lack is trustworthy relationships.

RDFknowledge graph
京ICP备2026014272号