Mental model

SkeinRank starts from a simple observation: technical teams rarely write the same concept the same way.

k8s, kube, kuber, kubernetes
pg, postgres, psql, postgresql
api server, api-server, kube-apiserver

Search systems can miss useful documents when these variants are not normalized. SkeinRank adds a deterministic terminology layer before or alongside retrieval.

Core idea

DictionaryCanonical values, aliases, slots, lifecycle statuses, stop lists.

MatcherFast local alias lookup with deterministic output.

AttributesStructured canonical values that can be attached to text or documents.

PassportTrace explaining where each attribute came from.

RetrievalSearch, rerank, evidence review, enrichment, or governance workflows.

Why not just embeddings?

Embeddings help with semantic similarity, but they do not always provide clear, stable, auditable terminology decisions. SkeinRank is not a replacement for embeddings. It is a complementary layer that gives retrieval systems more explicit signals.

Practical result

A messy sentence:

k8s rollout uses pg database

Can produce stable canonical values:

["kubernetes", "postgresql"]

And a reviewable trace that explains which alias produced which canonical term.