Search for cheukbaeksan (측백산) in Uibang Seongdo, and several records appear. Two use exactly the same Chinese spelling, 側柏散. Open their ingredient lists, though, and one names cheukbaekyeop where the other names baekyeop.
I kept those entries separate when building the site. A reader can put them side by side, see which names overlap, and follow each entry back to its source. Merging them under one convenient formula name would hide the very difference someone might have come to investigate.
Uibang Seongdo is a map for exploring public records about traditional medical formulas, medicinal materials, and diseases or symptoms. Records appear as stars, with lines that let you follow names mentioned in the source data.
Open the two records side by side ↗. Check the shared and differing ingredient names as you read how the comparison arrives at 50%.
Where the 50% comes from
The selected records come from the Korea Traditional Knowledge Portal, or KTKP. In KTKP, P0003343 lists Donguibogam (1613) as its source, and P0009496 lists Shiyi Dexiaofang (1345).
Both ingredient lists contain three names. Ginseng and hyeonggaesu occur in both. The remaining name is cheukbaekyeop in the first record and baekyeop in the second.
Across the two lists, that gives four distinct names, of which two are shared. The comparison therefore displays 2 / 4 = 50%. This calculation is the Jaccard index: the size of the overlap divided by the size of the combined set.
The number answers a narrow question about ingredient labels. It does not include doses or establish that the two formulas have half their effects in common. It also leaves the possible relationship between cheukbaekyeop and baekyeop open. Resolving that would take evidence beyond a name comparison.
I show the shared and exclusive names alongside the percentage because they explain what has been counted. A reader can see the two labels responsible for the difference without having to guess what “similarity” means here.
A name used for comparison is not a record ID
The comparison simplifies names to make the sets comparable. It standardizes character forms and removes some formatting, including the parenthetical part and trailing letter variants used in the source labels. The original labels remain available in the records.
That simplified comparison key has a limited job. I do not use it to decide that two source entries are the same record.
Drawing a line on the map calls for a stricter decision. A source field may mention another name without supplying its record ID. Before adding a link, the builder tries to resolve that name to one target, keeping the Korean name, Chinese spelling, and source variant together. If several records match, it leaves the relation unresolved instead of picking one.
The unresolved labels are counted separately for later inspection. The relation rules explain how the builder handles ambiguous matches.
Following the comparison back out
Open the two cheukbaeksan entries side by side, inspect their ingredient lists, then check the source details. Their IDs keep this pair identifiable even though the displayed names are identical.