Not Just

Data sources

Two layers power Not Just 个: the reviewed editorial data and an external research dataset used only for draft candidates.

Layer A: Reviewed editorial data

Every Published noun, sense, classifier, and usage is curated and reviewed by a Mandarin-speaking editor. The data lives in version-controlled JSON files in this repository and is rebuilt into a static search index on each release.

The reviewed layer includes:

  • Noun concepts with stable slugs.
  • One or more senses per concept, each with Simplified, Traditional, and lexical pinyin.
  • Classifier relations per sense: primary, common, contextual, or marked.
  • English explanations and pinyin for every primary phrase.
  • Reviewed examples for high-traffic entries.

Each entry records reviewed_at and reviewed_by. The reviewed_by field is currently a generic group name; that will be replaced with stable reviewer IDs in a later update.

Layer B: CC-CEDICT (planned, not active)

CC-CEDICT is a community-maintained Chinese-English dictionary distributed under CC BY-SA 4.0. We intend to use its CL: field as a wider fallback for nouns without a Reviewed entry. The snapshot dated 2026-08-24 will be used.

The fallback UI is currently disabled while we verify license and provenance. Until then, nouns without a Reviewed entry simply show "No reviewed answer yet".

License: Creative Commons Attribution-ShareAlike 4.0 International (CC BY-SA 4.0).

Attribution: CC-CEDICT, MDBG, mdbg.net.

Layer C: Public research dataset (draft candidates, active)

For Chinese nouns that have no Reviewed entry, Not Just 个 surfaces a draft answer sourced from the ChineseClassifierDataset (IJCNLP 2017, Peinelt & Liakata). It contains roughly 136,000 sentences with the correct classifier marked.

License: Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International (CC BY-NC-SA 4.0). Not Just 个 itself is free and offers Reviewed content under its own attribution. The dataset is not redistributed from this project; only the most-common-classifier lookup built from it is bundled.

Draft answers are clearly labeled in the UI as "Dataset match, not reviewed" and never produce indexable pages. They are removed automatically once a Reviewed entry for the same noun is added.

Layer D: Anonymous events (planned)

When feedback collection is enabled (by configuring NEXT_PUBLIC_EVENTS_ENDPOINT), anonymous events will be stored for content planning. We do not and will not collect account, name, email, or precise location.

See the Privacy page for the current state.

What is not in any layer

  • Realtime AI answers. The query path is deterministic.
  • Third-party tracking pixels.
  • User-generated content or public comments.