Work

A machine-readable corpus of the Wuzang Shanjing (449 mountains, 26 itineraries): text layer, interpretation ledger and evidence index

Year: 2026
Type: preprint
Abstract: The Wuzang Shanjing (五藏山經, "Five Storehouses: the Itineraries of Mountains"), the geographic core of the Shanhai Jing, is a Chinese text that enumerates 449 named mountains arranged in 26 itineraries ("chains"). Each chain is a directed walk: successive mountains are linked by statements giving a bearing and a distance in li, and each mountain carries a short description of its mineral produce, hydrology, flora, fauna and environment. We present interlinked, versioned, machine-readable files that cover the complete text: canonical_26_v61.json, a canonical base encoding all 449 mountains and 428 inter-mountain steps with character-offset pointers back to the source characters (the interpretation layer's direction-bearing scripts read canonical_26_v62.json, which differs only in seven direction tokens); 逐山落点台账.csv (Landing Ledger), a 449-row, 11-column ledger recording, for each mountain, the working interpretation attached to it, a confidence label, the load-bearing evidence, entries suspended pending further work, explicit counter-records, a status label, and the provenance of that interpretation; and an evidence index that decomposes the ledger's evidence column into 1,577 discrete, individually citable items. All files are keyed on the same (chain, intra-chain index) pair and were verified to agree on it 449/449. Two design choices make the corpus usable by researchers who disagree with the interpretation layer. First, the text layer is separated from the interpretation layer: the canonical base is derived mechanically from the source characters and is gated by a ten-item automated check, whereas every interpretive field lives in a second layer that can be replaced wholesale without touching the base. Second, discrepancies are recorded rather than reconciled: 23 of 26 chains show a numeric disagreement between the summary line the text itself gives at the end of a chain and the sum of the step distances parsed from the body of that chain, and these disagreements are preserved as data rather than silently corrected. This preprint is a data descriptor: it describes what the files contain, how they were produced and how they can be reused. It advances no geographical identification. The archived corpus, placement outputs and audit scripts are at https://doi.org/10.5281/zenodo.23909615 (v1.2.0; corpus CC BY-SA 4.0, code CC BY 4.0). This version corrects all descriptive figures against the v1.2.0 archive and supersedes preprint v1.0 (10.5281/zenodo.24286260).
Author: Ye, Lin
Language: English
Open Access status: green
OpenAlex ID: W7234035658
Cites:
Cited by:
Related to: 10