西藥略釋(1871;Bodleian Sinica 1798) / Xi yao lue shi [Brief explanation of Western medicines]: Transcription dataset
Authors/Creators
Description
《西藥略釋》(Xi yao lue shi)是嘉約翰(John Glasgow Kerr)編寫、1871年由羊城博濟醫局刊行的西藥學漢文著作;本底本為牛津Bodleian Libraries館藏Sinica 1798。全書含中英文序、藥物條目及中英藥名對照表,可供晚清醫學翻譯、藥名形成、傳教醫療與印刷史研究。本資料集以館方104幅影像的197個宣告範圍物理面為底本,提供逐面可定位的TXT、JSON、TEI XML、疑讀與版面部件資料。這是AI輔助轉錄修訂版,經獨立AI讀稿、逐面比對與回看原圖修訂;疑讀及未定碼位保留,未經人類全文雙校或專家審定,詳見Methods。
Xi yao lue shi [Brief explanation of Western medicines] is an 1871 Chinese materia medica by John Glasgow Kerr (嘉約翰), published by the Canton Hospital (博濟醫局) in Canton. This dataset transcribes the Bodleian Libraries copy, Sinica 1798, including Chinese and English prefaces, medicine entries and bilingual name tables, for research on late-Qing medical translation, pharmaceutical terminology, medical missions and print. It covers 197 targeted physical faces in 104 digital captures, with source-linked TXT, JSON, TEI XML, uncertainty and layout-part records. This AI-assisted revision follows a separate source-image reading, page-by-page comparison and checking against the source images. Unresolved readings and exact encodings remain marked. Full human double collation and expert review have not been performed; see Methods.
Notes
historical text and faithful mechanical transcription: No new exclusive restriction; original copyrightable documentation and arrangement: cc-by-4.0; source images excluded: Bodleian source image terms apply separately; CC BY-NC 4.0 is not a licence on this text-only package.. Source images are excluded. Rights in the historical text, original modern documentation/arrangement and library source images are distinct; source-image terms are not a second dataset-wide license. This fixed package was prepared before DOI archiving; current DOI links are supplied on the repository main page.
Methods
A separately sealed source-only AI reading was compared with fixed v0.1.0 without Unicode normalization. Source-reference tokens were treated as alignment atoms; differing reference IDs were not reading errors. AI nonblind adjudicators inspected source images for every declared difference and agreed high-risk regions. Only explicit local source-supported proposals were applied; unresolved glyphs, micro marks, point attribution and clipped folios remain. Original unit IDs and relative order are retained. Added small-attachment units are parallel annotations; a missing main column is inserted only at its explicit source-backed location. Native actual model, effort, counters and full native outputs are unknown or unavailable, not zero. No full human double collation, expert review, peer review or accuracy measurement was performed. Engineering checks and DOI do not certify readings. Source images/crops, dedicated code, prompts and complete private working records are excluded; a complete image-to-text reproduction method is not provided. Some initial-reading contexts were reused. The separately sealed source-only pass is recorded as not exposed to the initial transcription before sealing; hidden model context is unknown. Adjudicators then saw both drafts and are explicitly nonblind.
Files
bodleian-sinica1798-v0.2.0.zip
Files
(20.7 MB)
| Name | Size | |
|---|---|---|
|
md5:cc328367addbef97a382da37bb156233
|
1.5 MB | Preview Download |
|
md5:63003fa767532a8f08e802af9c88de6d
|
11.0 MB | Preview Download |
|
md5:e6c04228fd47b2c6572878718886ee1a
|
8.0 MB | Preview Download |
|
md5:900ed4614ba6cf220ea793d94219d83a
|
230.8 kB | Preview Download |
|
md5:31cebfef424d1599215e391f78861d77
|
2.3 kB | Download |