One Herb Connects to a Hundred-Plus Formulas

One Herb, Licorice, Connects to a Hundred-Plus Formulas

In another piece I wrote about why MateriaTrace is built to be source-traceable. This one is about something different: what makes the database itself good? Why would I call it the kind of knowledge base built for the AI era?

No adjectives. Just a few facts you can click and check.

One herb, licorice, connects to over a hundred formulas

Open the licorice (gan cao) page and you'll see something a flat dictionary can't do: it lists over a hundred formulas that contain licorice — and you can go from licorice straight into Gui Zhi Tang, then open each ingredient in that formula, and each of those leads on to the other formulas it appears in.

Cinnamon twig (gui zhi) connects to over sixty formulas the same way. Across the whole database, the "formula-to-ingredient" links alone number in the thousands.

That's the fundamental difference between a graph and a dictionary. A dictionary is a list of isolated entries — you look up licorice, and licorice has no relationship to other herbs or formulas. A graph is a web, where every node connects to others — and Chinese medicine knowledge is a web: one herb plays different roles in different formulas, one formula is built from several herbs in combination, and to understand any one of them you have to see where it sits in the network.

Building it as a network instead of a pile of isolated entries — that's the first advantage.

The scarcity is in the four things stacked together

On its own, a structured database isn't special. Neither is being source-traceable. MateriaTrace's value is in having four things at once — where most Chinese medicine information online has only one or two:

• Structured — every herb, formula, and concept is clean fields (nature, functions, composition, source…) that both machines and people can read precisely. Prose explainers can't do this.
• Source-traceable — every field traces word-for-word to the original classical text, with a checkable fingerprint. Most structured databases can't do this — they have data, but you don't know where it came from.
• Bilingual — Chinese classical text plus aligned English terminology; searchable and citable in either language.
• Networked — herb ↔ formula ↔ concept all connect, so you can drill down along the relationships instead of hitting a dead end after one lookup.

Prose explainers are readable but unstructured and unsourced. A typical herb database is structured but has no sourcing and isn't networked. The original classical books are traceable but slow to search, unstructured, with no links across entries. Doing all four at once is rare. That stack is the part that's genuinely hard to copy.

Why this is exactly the shape the AI era needs

Here's the key point: those four traits are exactly the shape AI can actually use.

When AI and search engines decide whether to trust a source — and whether to cite it — they increasingly care about a few things: is the content clearly structured, are the facts clearly sourced, can a machine read it. A sourceless prose explainer, AI won't cite (it can't verify it). A database with data but no provenance, AI can't trust either.

A knowledge graph that's structured, carries a primary source on every record, and connects its nodes to each other is the ideal raw material — it can pull "this herb, from this book, this line," it can follow the relationships to "which formulas use this herb," and every step checks out. In other words, MateriaTrace isn't only for people to look things up. Its very structure is built to be cited reliably in the AI era.

When AI can generate endless fluent-but-sourceless content about Chinese herbs in seconds, a structured, traceable, networked knowledge graph doesn't lose value — it gains it, because it offers exactly what generated content lacks: structure, provenance, and relationships.

Still growing

Right now the web holds over fifteen hundred herbs, several hundred formulas, and a set of pharmacological concepts, with thousands of links between the nodes — and it keeps growing: more public-domain classics, more formulas, more concepts, each new node added to the same standard (traced word-for-word, cross-checked, honest about uncertainty).

A network's value compounds as nodes are added — every new herb or formula brings more connections, and the whole graph gets denser and more useful.

Note: MateriaTrace records how the classical texts read and objective facts; it is not medical advice. Modern science's verdict on Chinese medicine is cautious and mixed (the NCCIH notes mixed results with no firm conclusions, and advises against using TCM to replace or delay conventional care).

Walk the web yourself

The best way to get it is to click through: open licorice and look at the hundred-plus formulas it connects to; pick one formula and look at its composition; then open another herb from that composition and see where it leads. A few steps in, you'll feel it — this isn't a dictionary you close after one lookup. It's a web that gets deeper the further you walk.

If this "everything is connected, and all of it clicks back to the original" approach is useful to you, see what's in MateriaTrace on the homepage.

MateriaTrace is a source-traceable, bilingual knowledge graph of Chinese medicine. It records how the classical texts describe each entry, with every field traceable to the primary source, and makes no treatment or usage recommendations. Every figure in this article can be checked on its entry page.