Munjip Metrics
한국어

Method and limits

Sources

Institute for the Translation of Korean Classics, 'Hanguk munjip chonggan' (data.go.kr dataset 3074298, no license restriction): 1,249 XML archives adopted from the 1,251 portal entries (one without an attachment and one old-schema edition excluded), one per collection — a bibliographic file (author, series volume, publication year, table of contents, short introduction) plus one file per fascicle with every work's title, writer, genre class and text. Counted at the work level (type="최종정보").

Academy of Korean Studies, 'Successful candidates of the civil service examination' (15052752): 15,151 passers with name, courtesy and pen names, clan seat, birth year, examination year and grade. Academy of Korean Studies, 'Korean historical persons': 27,035 persons with names, courtesy and pen names, dates and clan seat; used to cross-check author dates.

Unit of counting

Every number is counted per work: one level element of type="최종정보" in the ITKC XML. Of 637,203 works, 609,494 are by the collection's author (writer name equals the bibliographic author). Characters are CJK code points inside <내용>; those inside original, collation and marginal notes are counted separately as note characters. Check: for all 1,249 collections the count equals an independent recount of every <내용> in the raw XML; zero residual character references.

Genres

The genre classification the ITKC attached to each work (7 major, 28 minor classes) is used as given; nothing was corrected or added. Classification quality is shown by the residual 'other' share (B7) and by a blind coding of a 300-work sample reported on the reliability card (marked as in progress where not yet done in 1.0).

Centuries

The XML carries no date per work, so every time series is grouped by the author's birth century. Printing lags the author's active years by about a century (Kwak 2026; publication lag D3). Collections whose author has no birth year sit in a 'birth year unknown' row.

Merging authors

The author id is the Chinese-character name plus the birth year. Several collections by one author (main collection, continuation, posthumous works) are summed in the author table with the number of collections shown. Works whose writer differs from the author (prefaces, postscripts, biographies, elegies, appendices) are excluded from the author's output. Two different people sharing a name and birth year cannot be told apart; such cases are corrected through overrides when found.

Person linkage

Examination-roster rules: M1 same name and birth year (first pass when a person is listed for re-examinations), M2 same name, roster birth absent, courtesy or pen name identical, M3 same name, a single candidate, examination year within the lifetime. Where more than one candidate survives the author stays unlinked (AMB). Linkage is not a pass rate. The AKS person database is matched by name and birth year (P1) or name and pen name (P2) and used only to check death years.

Relationship genres

A letter's addressee is the string after an opening 答·與·上·寄·復·報·奉·呈·贈·簡·謝 up to 書·簡·問目·別紙·狀 etc.; an epitaph's subject is the string before 墓誌銘·墓碣銘·神道碑·行狀 etc. with honorific prefixes removed. Version 1.0 does not merge strings into persons: the same person named by courtesy name, pen name or office counts as different addressees. Parsing precision is measured on a sample.

Known limits

Reproduction

The GitHub repository's src/ holds the fetcher (fetch_itkc.py), work table (build_works.py), checks (check_works.py), author linkage (link_authors.py), metrics (metrics_a–d.py), reliability (reliability.py) and site (build_site.py). The portal lists 1,251 entries; 1,250 attachments were downloaded (one entry has none) and 1,249 archives adopted (one old-schema duplicate excluded); their SHA-256 digests are in raw_manifest.json.

Romanization

Collection titles and personal names follow the Revised Romanization of Korean (2000); the surnames 김 and 이 are given in their customary forms, Kim and Yi.

Versions