Chapter 6. Two Classification Criteria — The Nine Categories and Worldview Axes

What happens when an IP has only one classification criterion? Open the character category of “Gongdong” and association executives, dungeon-raiding awakened people, and ordinary citizens all sit together. All three are characters. Yet character classification alone cannot tell us whether each belongs to the association domain, dungeon domain, or everyday-life domain. The same limitation appears in every IP.

The demonstration IP frequently used in this book is “Gyunsahae”. It is an SF ecological world in which a mycelial network covering the planet is the medium of magic, memory, and communication, and species divide according to the fungus with which they symbiose. Follow this one IP through the end to see what orthogonality means, and how the two classification criteria mesh becomes clear.

Librarying uses two classification criteria: the nine categories and worldview axes. The two criteria are orthogonal. The nine categories ask, “What function does this keyword serve in narrative?” Worldview axes ask, “To what structure of our worldview does this keyword belong?” Each headword has one answer to each question; together, the two answers are its coordinates.

If There Were Only One Classification Criterion

See what happens when only the nine categories exist through “Seongra Gangho,” a martial world where martial arts are bound to constellations. Open the character category and you find an orthodox sect leader, an unorthodox assassin, a physiognomist who reads celestial patterns, and a wanderer roaming the martial world. All four are certainly characters in the sense that they function as agents in the story.

Yet character classification alone makes it difficult to know how these four relate or what domains they occupy in the worldview. The orthodox leader spans the orthodox–unorthodox faction domain and the sect domain; the physiognomist touches the celestial-pattern domain where martial arts bind to stars; and the wanderer belongs to the martial-world domain, outside every sect. Though all are “characters,” they distribute across wholly different worldview domains.

If this distinction is not expressed in the Vault, a search for “show every character in the celestial-pattern domain” cannot be answered; one must open the entire character category and choose each by eye. With 1,500 headwords, that is burdensome. Adding worldview axes as a second classification criterion solves exactly this problem.

Faceted Classification: The History of Two Independent Criteria

Using two independent classification criteria together is not Librarying’s invention. Information science and library science have a methodology called faceted classification.

Traditional hierarchical classification descends by one criterion, as in animal → vertebrate → mammal → carnivore → feline. It is intuitive, but limited to one criterion, and cannot make a search that simultaneously applies region and taxon, such as “mammals living in Africa.”

Faceted classification differs by dividing one object into several independent facets, or faces. For one book, its subject may be history, region Korea, period Joseon, and form novel. Since facets attach independently, a reader can search by any one of them or combine all four, as in “Korean historical novel of the Joseon period.”

Obsidian’s tag system implements faceted classification, and Librarying uses the nine categories and worldview axes as two independent facets within that system.

What It Means to Be Orthogonal

In mathematics, two axes being orthogonal means they operate independently: a value on one does not determine a value on the other. If the X-axis decides “which direction,” and the Y-axis “how high,” they are orthogonal: something pointing east may be low or high, and something high may point east or west.

See the orthogonality of the nine categories and worldview axes through “Gyunsahae.” Suppose its worldview axes are six: mycelial network, species, symbiosis, status, terrain, and general. “Light-fungus folk” is a character and species; “empty spore sac” is an object and species; “symbiosis” is a concept and symbiosis; “mycelial connection” and “shared memory” are mechanisms and mycelial network; “ruler holding the great node” is a character and status; “barren land” is a place and terrain; and “luminous mycelium” is mise-en-scène and terrain. The same worldview axis, species, contains characters, objects, and concepts; the same narrative category, characters, contains the light-fungus folk on the species axis and the ruler on the status axis. Each headword is given coordinates as one nine-category value times one worldview-axis value. The fact that neither coordinate determines the other is orthogonality.

When Events Bind People, Places, and Objects

One perspective is often missed when designing worldview axes: separating people, places, objects, and events is not enough to reveal their relations. This is why ontologies such as CIDOC CRM, which integrate cultural-heritage information, do not treat people, places, objects, and time as merely independent lists; they seek to connect them through events and actions. A museum object gains meaning when the era, place, hands, and event through which it passed are recorded together.

An IP is similar. A sword is an object, a swordsman is a character, and a martial arena is a place. But what binds them is the event: from whom the sword passed to whom, in which arena it broke, and how the sect’s authority changed after that event. Without events, people, places, and objects remain separate lists; when events enter, three lists connect into one memory of a world.

Worldview axes therefore are not simply “character axis,” “place axis,” and “object axis.” The nine categories already separate characters, places, and objects. Worldview axes must show the worldview domains and event flows in which they meet. In one IP, “succession to the throne” becomes an axis; in another, “mycelial network”; in another, “enshrinement.” A good axis is not a name that splits a list more finely, but a name that explains a scene in which people, places, objects, and events come together.

Why Worldview Axes Are Needed

With only the nine categories, opening characters mixes protagonists and minor roles, people from different worldview domains. Their narrative function is the same—character—but within the IP they belong to entirely different domains.

In Vault graph view, character-category points cluster in one color. To determine which worldview domain these characters occupy, another color distinction is needed: the worldview axis. Open a Vault and graph view in Obsidian and points can be colored in two ways. By the nine categories, characters are blue, places green, and mechanisms orange, separating narrative functions. By worldview axis, each domain receives a different color, separating worldview areas. The same Vault is viewed through two lenses. With these two lenses, searches such as “show every keyword in this particular nine category within this particular worldview axis” become possible.

Worldview Axes Differ by IP

The nine categories apply equally to every IP because this book defines them. Worldview axes differ: they are designed for each IP.

Consider five IPs. Each establishes five core domains plus “general” and “undecided,” for seven axes. The modern hunter IP “Gongdong” uses dungeon, awakening, association, concentration, and everyday life. The romance fantasy “Hyanggung”, an empire where scent mediates status and magic, uses scent, status, social world, family, and divinity. The martial-arts “Seongra Gangho,” where martial arts bind to constellations, uses martial arts, celestial patterns, orthodox–unorthodox factions, martial world, and sects. The Eastern-dynasty story “Usabu”, where rain proves legitimacy, uses heavenly mandate, water, offices, powerful families, and public sentiment. The space SF “Sangyeoho”, an empire where the dead guide routes, uses road of the dead, navigation, family, enshrinement, and dimension. The five IPs’ core domains are visibly all different.

Now observe where one object with the same nine-category function—an object, like a staff—goes in different IPs. In “Seongra Gangho,” a natal-star tablet engraved with a constellation determines a martial artist’s star, so it belongs to martial arts. In “Hyanggung,” a perfume bottle holding the scent worn that day acts as an identity card, so it belongs to status. In “Gongdong,” a rank card separating an awakened person’s status is a marker issued by the association, so it belongs to association. In “Sangyeoho,” a spirit tablet fixing the dead is central to the navigator’s navigation, so it belongs to navigation. Narrative function remains object, but worldview structure places it on martial arts, status, association, or navigation.

Characters likewise fall on different axes by IP. The grand aristocratic perfumer of “Hyanggung” belongs to the social-world axis; the chief scholar of Usabu, head of an office that speaks for heaven, belongs to offices; the enshrinement-house priest of “Sangyeoho,” who governs death, belongs to enshrinement. The nine-category function remains character, but its axis changes with the IP’s domains. There is no one correct worldview-axis pattern. Even two fantasy IPs differ if one centers magic and the other politics, so first identify what the IP centers and then design its axes.

The table of contents of the Forgotten Realms Campaign Setting expresses this axis idea directly as chapter structure. Worldview domains such as Magic, Deities, History, Organizations, and Life in Faerûn are divided into chapters, and each chapter contains characters, places, objects, and events together. One character often appears in both the Deities and Organizations chapters. Thus this chapter structure corresponds to worldview axes, not the nine categories. The Eberron Campaign Setting works the same way, differing only in the names and number of domains.

D&D’s division of worldviews by domain forty years ago was still only one classification. Dividing by domain gathers characters, places, and objects within that domain, but does not tell whether a character is an agent in narrative or an object is a narrative mechanism. Librarying borrows domain classification as worldview axes and overlays the nine categories orthogonally. Only this intersection of two axes enables searches such as “only character headwords in the faith domain.” Where conventional gazetteers and personal wikis usually divide materials by one criterion, Librarying reads IP material through two overlapping criteria.

Principles for Designing Worldview Axes

Worldview-axis design has four principles.

Keep the number to five through seven. Too few—two or three—makes classification so coarse that it loses meaning. Too many—ten or more—makes management difficult. Compressing an IP’s core worldview domains into five to seven is the recommended range.

Always include “general” and “undecided” axes. “General” classifies vocabulary not belonging to a specific worldview domain: words such as “hello,” “yesterday,” and “big,” usable independently of the IP. “Undecided” is a temporary classification for headwords whose axis has not yet been determined. Without these two, unclassifiable headwords remain nowhere until the review stage reveals them.

Decide inductively. Rather than first constructing axes in one’s head and forcing headwords into them, scan one hundred headwords first and discover which clusters arise naturally. This posture creates worldview axes suited to the IP and is the core work of profiling.

Do not freeze axes, but do not change them casually. Adjustment may be needed while second integration proceeds. If many headwords accumulate under “general,” a new axis can emerge from them. But every new or changed axis must be recorded in the profiling file, and the axes of already assigned headwords must be checked again. Worldview axes function as permitted values in the second-stage CSV and frontmatter, so a name slipping in from outside the list is caught as an error during verification.

Why They Must Be Decided in Profiling

Chapter 3’s time budget placed “profiling” first because second integration becomes difficult if worldview axes are not fixed in advance.

After first extraction, when the 5,000 to 15,000 rows in the five-column CSV are given to an AI work partner for thirteen-column integration, the worldview-axis column needs a list of permitted values. Without it, AI attaches axes improvised headword by headword, and names vary across 1,000 entries. In “Gyunsahae,” the same axis may split into “mycelial network,” “mycelium,” “mycelial-network worldview,” and “mycelial realm.” Fix the allowed worldview-axis list during profiling and deliver it with the second-integration task; the AI work partner then attaches only axes inside that list. Notation stays unified.

The profiling output is a file named {project}_profile.md. Along with permitted worldview-axis values, it contains IP-specific notes for the nine categories, a list of supporting reference documents, and tone guidance for fourth-stage narrative writing. This file is the starting point of a Librarying first cycle.

Inductive Decisions: Let Headwords Speak Their Axes

Use “Gyunsahae” to see the concrete posture for deciding worldview axes in profiling. Before first extraction, or after previewing part of its results, you read the IP’s documents and begin to notice vocabulary clusters that repeatedly occur together on a page. In every document, mycelial network, node, connection, and shared memory appear together. Elsewhere light-fungus folk, rot-fungus folk, stone-fungus folk, and spore wanderers appear together. In another place light-fungus forest, rot-fungus swamp, stone-fungus mountains, and barren land appear together.

These clusters are candidates for worldview axes: mycelial network, species, and terrain. If the cluster in which a person combines with a fungus determines ability, it becomes symbiosis. If power gathers to someone holding a great node, it becomes status. General is for vocabulary that enters none of these clusters. Add “undecided” for headwords not yet entering anywhere, and these seven axes are fixed in profiling. During second integration after first extraction, each headword is placed on one of the axes.

This inductive posture lets headwords reveal their own axes. Rather than assigning axis names first, inspect the domains into which the IP is actually divided. Put names first and you force headwords into them; discover patterns first and headwords naturally sit in their categories.

When Two Coordinates Exist Together

With both classification criteria, the Vault’s information structure becomes dimensional.

When handling an episode in “Gyunsahae” that describes the operating principles of the mycelial network in detail, you may want all related mechanism headwords. Apply worldview axis “mycelial network” and nine category “mechanism” together, and every mechanism headword in that domain appears; it becomes immediately clear how connection and shared memory mesh. The same search works in other IPs. To write a “Sangyeoho” short story about navigation and collect mechanisms from that domain, multiply worldview axis “navigation” by nine category “mechanism.” To select characters for a social-world scene in “Hyanggung,” multiply “social world” by “character.” Whatever the IP, multiplying worldview axes and the nine categories produces the desired intersection.

Making this two-coordinate search possible is the practical reason to introduce worldview axes. With only one criterion, the only method would be opening every character and checking one by one. When two criteria are orthogonal, the desired intersection can be drawn immediately.

The Actual Shape of Frontmatter

When third skeleton building creates an .md file, its frontmatter looks like this:

---
title: Mycelial Connection
aliases: [connection, mycelial-network connection]
category: mechanism
worldbuilding_axis: mycelial network
english: Mycelial Linking
romanization: Gyunsa Jeopsok
version: current
tags:
  - category/mechanism
  - axis/mycelial-network
  - version/current
sources:
  - Rulebook 3rd edition p.24
related:
  - "[[Shared Memory]]"
  - "[[Aseptic Person]]"
---

The category field is the nine-category value, and worldbuilding_axis is the worldview axis. These fields enter the tags array as category/mechanism and axis/mycelial-network, making them searchable. Chapter 7 revisits this frontmatter structure when it treats third skeleton building in detail.

Two Pitfalls in Operating Worldview Axes

Two common pitfalls deserve notice. First is worldview-axis inflation. An IP may begin with six axes, but if a new axis is created whenever a headword feels awkward in “general,” it can soon grow to twelve. Beyond ten axes, color distinctions lose meaning in Vault graph view, search filters become complex, and management is burdensome. When a new axis seems necessary, first see whether an existing axis can absorb it. Create one only when absorption is impossible, and record every change in the profiling update.

Second is confusing nine categories with worldview axes. In “Gyunsahae,” this happens when someone confuses worldview axis “mycelial network” with nine category “mechanism,” puts every mycelial-network keyword in mechanisms, or tries to make “mechanism” a worldview axis. The nine categories ask narrative function—what something does—while worldview axes ask worldview structure—where something belongs. One mycelial-network domain therefore contains mechanisms, characters, values, and events. Keep the orthogonality of the two criteria in mind.

Supporting Material for Hub Decisions

Chapter 8 treats hub decisions in detail, so here we only note how worldview axes support them.

Within a worldview axis, the headword with the highest related-keyword citation count is a hub candidate for that axis. In “Gyunsahae,” the most-cited headword in the mycelial-network axis is that axis’s hub candidate, and the most-cited one in the species axis is the species-axis candidate. Finding hubs by axis reveals the core headwords in every worldview domain. Place them alongside the top headwords by overall citation count, and the Vault’s structural center emerges in three dimensions.

Only these two lenses make certain questions visible. Comparing “What are the hub headwords in the mycelial-network axis?” with “What are the hub headwords in the species axis?” shows which axis has more hubs, and therefore in which worldview domain the IP has developed most deeply. If five years of operation produce forty hubs in the mycelial-network axis but five in species, it signals that the species worldview is relatively underdeveloped. That signal becomes material for deciding the next creative direction.

Holding the Two Coordinate Systems

The permitted-value list fixed during profiling is delivered with the second-integration task. AI work partners attach candidate axes to each headword, and people confirm them in review. What one classification criterion cannot reveal becomes visible through intersection searches across two criteria.

Chapter 7 is third skeleton building: once the thirteen-column CSV is ready, it is automatically converted into 1,500 Obsidian .md files. Human hands almost do not touch this automation stage, but its result finally spreads 1,500 points and lines across the screen.