How the Japanese name reference is built

1. Names begin with recorded forms

Given-name and surname pages begin with JMnedict entries. Matching romanized readings are grouped for browsing while each recorded written form and kana reading remains visible.

2. Kanji meanings stay at character level

KANJIDIC2 English glosses describe individual characters. A multi-kanji spelling is not automatically rewritten as one definitive English sentence, because dictionary glosses alone do not establish the intention behind a particular person’s name.

3. Character-set checks are references, not registration advice

KANJIDIC2 grade values help distinguish jōyō and jinmeiyō classes. The Ministry of Justice remains the official reference for characters usable in children’s names, while kana spellings are handled separately.

4. Coverage varies by name

Some readings have many recorded written forms and rich kanji context; others have only a small number of dictionary records. The page shows the available evidence rather than filling those gaps with invented detail.

5. Updates preserve completed data

When the underlying dictionaries are refreshed, readers continue to see the last completed reference set until the replacement has been fully processed. This avoids showing a partially updated name collection.

6. Romanization is a reading aid

Name pages generate a consistent display from the source kana reading using the current Japanese government romanization guidance as the reference, including long-vowel marks, current small-ッ consonant doubling, and disambiguating n’ before vowel or y sounds where the kana requires it. Search also accepts conservative romanization variants and normalizes hiragana/katakana input. Kana alone does not always establish morpheme/vowel boundaries, so the page also keeps the source kana and a vowel-sequence form visible. This does not establish a particular person’s preferred Latin-letter spelling, and pitch accent is not inferred.

7. Browser audio is convenience TTS

The pronunciation button uses Japanese text-to-speech available on the visitor’s browser or device. It is not presented as a native-speaker recording or pitch-accent source.

8. Comparison is local and evidence-preserving

The comparison workspace stores only the selected source-recorded forms in the visitor’s browser. It compares readings, character-level glosses, stroke totals and loaded character classifications without creating permanent comparison pages or inventing a whole-name translation.

9. How kanji-to-name discovery works

Character pages retrieve given-name and surname groups from the completed JMnedict reference set using exact character-to-name relationships. Corpus counts and ordering describe the imported dictionary data, not population prevalence or popularity.

10. Relationship links are literal, not semantic guesses

Shared-kanji links require an exact character overlap in source-recorded spellings. Same-writing links require an exact written form with another source-recorded reading. Neither relationship is presented as proof that two names have the same origin, meaning, gender or cultural usage.

11. Explorer filters are temporary research views

The Name Explorer combines source-recorded name type, dictionary gender classification, opening reading family, mora count, writing type and up to two exact kanji. Both selected characters must occur in the same recorded spelling. Filter combinations remain temporary research views; name and kanji reference pages are the permanent destinations.

12. Kanji-pair research is co-occurrence, not popularity

The pair tool checks exact JMnedict written forms. One-character discovery highlights partners only after at least two distinct recorded forms, while explicitly searched sparse pairs remain visible with a warning. Pair counts describe dictionary-corpus co-occurrence only and are not population popularity, etymology or a claim that two characters form a traditional semantic unit.

13. Spelling anatomy stays descriptive

Name pages summarize reading distribution, stroke-count coverage, loaded character classes and literal kanji positions across their source-recorded forms. Kanji pages can summarize whether a character occurs alone, first, final or internally in published JMnedict records. These structural counts describe the imported corpus only; they are not style ratings, popularity measures or claims about naming intent.

14. Accessibility and stable reference URLs are part of publication quality

Public pages provide keyboard-visible focus, a skip-to-content link, horizontally reachable mobile navigation and scrollable-table regions that remain keyboard accessible. Paginated reference collections receive page-specific titles and descriptions, while search and research tools remain temporary discovery views rather than a second permanent reference library.

15. Main reference pages need measurable source depth

A name or surname can remain searchable even when its source footprint is small. Inclusion in the main reference library uses group-level evidence derived from JMnedict: a group must meet the basic source-quality checks and then have either at least three distinct recorded spellings, or at least two spellings backed by four or more spelling/reading source rows. More limited pages can remain directly accessible without being promoted into the main library; the rule is about source depth, not word count.

16. Coverage labels are separate from library-inclusion rules

Public name pages describe their source coverage in reader terms such as standard, extended or limited. Library inclusion thresholds are kept separate from those reader-facing coverage labels.

17. Source refreshes are compared before publishing

When a changed JMnedict source completes, the previous and replacement reference sets are compared before the earlier set is retired. The refresh process records meaningful additions, removals and changes in the source-backed name groups before the replacement is published. These deltas describe changes in the imported reference corpus and reference-library state, not changes in real-world name popularity.

18. Source verification and data age are tracked separately

Each source check records when the reference was verified and whether the underlying download changed. Verification recency is tracked separately from the age of the published source material, because an older reference can still have been freshly rechecked.

19. Failed refreshes preserve the last completed corpus

Source health tracks consecutive failures and recovery without replacing completed public data with partial work. JMnedict and KANJIDIC2 replacements are prepared separately from the currently published reference and become public only after the replacement dataset completes its integrity checks. A failed refresh does not replace completed public data, and other independent authority sources can still be checked.

20. Reading ambiguity and kanji roles stay literal

Name pages can summarize whether an exact written form has multiple source-recorded readings and how recurring kanji appear across the displayed spellings. Alternative-reading counts come from exact JMnedict writing matches. Character-role tables use only literal first, final, internal or one-character positions plus KANJIDIC2 character metadata; nanori are shown as character-level reference data and are not used to reverse-engineer a whole-name reading.

21. Kanji pages separate character readings from full-name readings

Kanji pages may show complete JMnedict name forms and their complete kana readings, writing-length profiles and literal co-occurring characters. A full-name reading is never treated as proof of how one particular kanji is pronounced inside that multi-character spelling. Co-occurrence highlights require at least two distinct recorded written forms and remain corpus relationships rather than popularity or semantic-compound claims.

22. Exact-writing navigation stays literal

Name pages connect an exact written form to all source-recorded whole-name readings, its constituent kanji reference pages and, for bounded 2–4 kanji forms, other public JMnedict forms containing the same contiguous character sequence. Sequence candidates are checked against exact source-recorded character sequences; a shared sequence is not presented as proof of shared etymology, meaning or cultural relationship.

23. JMnedict source text stays attached to its own classification block

JMnedict trans_det values are imported only from the same translation block that supplies the recognized given-name or surname classification. Source text that merely duplicates the generated romanization is omitted from the public page; distinct source labels or descriptors can be shown with an explicit boundary that they are not inferred kanji meanings, popularity evidence, or proof of a person’s preferred Latin spelling.