Repository path: workshop/translations/kongyiji/provenance.md · rendered 2026-09-09

〈孔乙己〉 — source text and the six-segment selection

Lu Xun 魯迅 (1881–1936), 〈孔乙己〉, first published in 《新青年》 卷六 (March 1919), collected in 《吶喊》. Public domain: the author died in 1936, and the witness used here is the 1938 《魯迅全集》 text, whose Wikisource record carries both {{Pd/1996|1936}} and {{PD-China-Organization}}.

Witness

Chinese Wikisource, 孔乙己, fetched via the MediaWiki action=parse API on 2026-08-07. Traditional characters, 『』 quotation, the 1938 《魯迅全集》 orthography (豫備, 儍, 喫, 于, 靑, 淸, 偸, 煑, 纔). Rendered from a Proofread <pages index="魯迅全集01_(1948).pdf" from="313" to="318"/> transclusion; the page-break zero-width markers were stripped and paragraphs reassembled from the <p> elements. Thirteen paragraphs, 2,601 characters in source-zh-full.txt.

One witness only. Unlike wang-liulang, no second witness was collated: no experiment here turns on a textual variant, every segment is modern vernacular with no transmission history, and collation would have been instrument work inside a unit that does not need it. This is a stated limitation, not an oversight — a variant in the 1938 text would be invisible to this project.

The selection

Six segments, 1,201 characters, source-zh-segments.txt. They are the story's first half up to the bean-dish scene, minus three paragraphs (the narrator's employment ¶1, his boredom at the counter ¶2, Kong Yiji's backstory ¶4) and minus everything after the beans (the broken legs, the last visit, the death).

Why this story. The selection was made for one property, declared before any rendering existed: Kong Yiji's speech is built out of classical-Chinese material — the particles 之乎者也, the Analects tags 君子固窮 (XV.2) and 多乎哉？不多也 (IX.6), the classical 竊 against the vernacular 偸, the examination word 秀才, the character 回 and its four graphs — every one of which is a place where a translator must choose between carrying what the source has and writing English a reader can move through. That is the fork ARM-sense-tradeoff exists to measure, and it is denser here than anywhere else the project has translated.

Prior renderings by this project

None. Checked mechanically across the whole repository before selection (孔乙己, Kong Yiji, 咸亨): zero hits. This check exists because S128 discovered it had re-rendered a passage the project had already translated eleven days earlier and had never asked (note (bhb), and RS-20260807d limit 8). It is now run before the source is chosen, not after the run is over.

The project has translated Chinese three times before — wang-liulang (Pu Songling, classical), yanfu-yili-yan (嚴復 1898, classical), yingyi-jiejixing (modern critical prose). This is its first Chinese literary fiction and its first Lu Xun.