Background · prior art survey

Has this been done before?

A survey of prior work behind the idea of reading philosophy like code — keeping the exact prose intact while overlaying logical-role color, nested typed brackets, and clickable concept-entities with coreference, find-all-references, jump-to-definition, and identity relations. Compiled from three parallel research passes across HCI / augmented-reading, tools-for-thought / notation, and argument-mapping literatures.

The short answer

well-troddenNearly every individual ingredient has deep precedent. Typed bracketing is Montague semantics & Frege made visual, and is what projectional editors (Hazel, MPS) do for code. Concept-entities + backlinks is Xanadu transclusion, Roam/Obsidian, NLP coreference — and, almost exactly, legal "defined-term" linkers. Identity relations are Peirce's lines of identity and DRT's discourse referents. The inferential structure of philosophy is the entire argument-mapping field.

freshThe synthesis is unattested: no tool found colors natural-language connectives/quantifiers/modals by logical role, and almost all close cousins either author in a restricted notation or extract the argument into a separate node-link diagram that discards the prose. Keeping the exact text and overlaying Montague-style typing + DRT-style coreference + IDE affordances in situ, as a reading instrument, is the original bet.

A Historical notations for the logical skeleton

the deep ancestors — drawing the structure of a judgment

Frege — Begriffsschrift · 1879 · SEP2-D “concept-script” drawing negation, conditional & quantification off a content stroke; replaces subject/predicate with function/argument. The origin of the idea your bracketing enacts — a judgment as a functor saturated by typed operands.

Montague Grammar / Semantics · 1970–73 · SEPEvery constituent gets a semantic type; meanings combine by function application over typed lambda calculus. Your typed bracketing (frame + operands) is Montagovian function application made visual.

Discourse Representation Theory · Kamp 1981 · SEPDiscourse as nested boxes of persistent discourse referents that corefer across sentences. The direct ancestor of your cross-text concept-tracking; the box is its bracketed frame.

Peirce — Existential Graphs · c.1896 · SEPDiagrammatic logic where one individual is a continuous line of identity — prefiguring your identity links between two phrasings of the same concept.

Reed–Kellogg sentence diagrams · 1877 · wikiBracketing a sentence into nested labeled constituents on a visual frame — the literal analog ancestor of your decomposition, and the precursor of modern parse trees / displaCy.

B Argument mapping & CSAV

the field that most directly visualizes the logic of philosophical prose — but as a separate diagram

Araucaria · Reed & Rowe, 2001 · arg.techSelect a prose span, tag its argumentative role in place (stored as Argument Markup Language). Exactly your move — except it renders to a node diagram; you render the markup back as colored/bracketed prose.

Rationale · van Gelder / Austhink, 2006 · reasoninglabArgument maps with colored/shaped nodes (green reasons, red objections). The visual ancestor of your role-coloring; you inline the roles into running prose instead of extracting boxes.

Carneades · Gordon, Fraunhofer · github.ioTyped argument graphs with argumentation schemes & statement-level acceptability — mirrors treating each concept as a first-class resolvable node.

Toulmin model · The Uses of Argument, 1958Six labeled roles (claim, data, warrant, backing, qualifier, rebuttal) — the conceptual vocabulary behind annotating which clause is claim vs. warrant.

Kialo · kialo.com · Visualizing Argumentation · Kirschner et al., 2003Kialo is a coarse sentence-level pro/con tree; the Springer volume coined “CSAV.” Your twist: keep the visualization embedded in the prose rather than replacing it. caveat: van Gelder's ~0.8 SD gains are largely his own group's evidence

C Augmented / active reading interfaces

the closest spiritual ancestors for gloss + cross-reference over a document

ScholarPhi · Head et al., CHI 2021 · arXivClick a term/symbol for an in-context definition pulled from elsewhere in the doc, plus “show every use.” Almost exactly your gloss + find-all-references + jump-to-definition — but for nonce terms, with no logical typing or coreference.

Semantic Reader Project · Allen AI / UW, 2023 · arXivUmbrella effort for AI-augmented reading of scholarly PDFs — the canonical “augmented reading” program, aimed at citations/terms rather than logical structure.

Potluck · Ink & Switch, 2022 · inkandswitchNearest mechanism-level precedent: recognizes structured entities inside freeform prose and attaches live behavior. You specialize this to logical roles & coreference. design warning: they dropped manual highlighting — it created “hidden state outside the text”

Explorable Explanations · Bret Victor, 2011 · worrydream · Threddy · UIST 2022Victor's charter for “active reading”; Threddy is highlight-to-sidebar entity clipping. Both share your interaction shape; neither touches latent logical structure.

D Tools for thought · projectional editors · notation

the “like code / type system” lineage

Hazel · POPL 2017 · hazel.orgA live environment over a bidirectionally typed calculus with “typed holes” — the closest formal precedent for “frame functor + typed operands,” where a hole is an operator awaiting typed arguments and the tree is always well-formed.

JetBrains MPS · Lamdu · Dion · projectional editingManipulate a typed AST directly and project it as surface notation. Existence proofs of your architecture — structure as source of truth, brackets/prose as projection — but for code.

Iverson — “Notation as a Tool of Thought” · Turing lecture, 1979 · pdfThe foundational citation for the thesis that a well-designed notation changes how one reasons, not just how things are stored.

Matuschak & Nielsen — Transformative Tools for Thought · 2019 · numinous · Subtext · BranderThe manifesto your slogan answers to; Subtext is the prose-side “notation as a tool of thought,” decomposing text into typed blocks.

Ink & Switch — Embark & Crosscut · 2023 · inkandswitchEmbark's typed “mentions” = your clickable concept-entities; Crosscut's “meta ink” declares equality/identity between elements — the closest analogue to your typed identity relations (on sketches, and deliberately avoiding a type system).

E Coloring by role — grammar & logic

the least-developed thread for prose, and the one you're mostly inventing

Montessori grammar symbols · refA fixed shape+color per part of speech (noun = black triangle, verb = red circle…). The closest historical precedent for a color-per-function legend; you generalize from parts of speech to logical roles.

Colourful Semantics · SLP / literacy tradition · POS prose highlighters · natural_syntaxColor grammatical roles (blue subject, red verb) to reduce load. You go a level deeper — coloring logical/semantic role, not surface syntax. efficacy is clinical/anecdotal, not RCT-backed

Syntax-highlighting evidence · Sarkar, PPIG 2015 · pdfEye-tracking found highlighting sped comprehension (~8s), shrinking with expertise — the closest peer-reviewed support. But it's about highlighting broadly; no study isolates role- or bracket-coloring. there is also a “case against syntax highlighting,” and Bionic Reading failed its evaluations

F Rainbow brackets & code affordances for text

rainbow-delimiters / VS Code Bracket Pair Colorization · 2021Depth-colored brackets, standard in every major editor; VS Code's incremental (2,3)-tree took a 42k-line update from >10s to <1ms — proven-cheap algorithms you can borrow directly. the Fahlman ~1981 origin story is folklore, not primary-sourced

Pad++ / semantic zoom · UIST 1994 · code folding & outline viewsRender simplified abstractions when zoomed out — the UI ancestor of folding a paragraph to its thesis (“zooming out” on argument structure).

G Concept linking — backlinks & defined terms

your entity layer is essentially a solved product in two adjacent domains

Roam / Obsidian backlinks · 2019 · roam · Ted Nelson — Xanadu · transclusion, 1960s“Unlinked References” = find-all-references; [[link]] = jump-to-definition for a named concept; transclusion is the deeper hypertext ancestor. Applying LSP-style go-to-definition to prose concepts specifically is still largely open. Roam “40% faster” claims are unverified

Contract Tools · Definely Draft · papersoftwareThe single closest shipping implementation of your entity feature: double-click a defined term to jump to its definition & highlight every instance; Definely auto-generates a “Definitions Report” (≈ your references panel) — just for contracts. a named “Legal Begriffsschrift” project appears not to exist

Coreference resolution + entity linking · Jurafsky & Martin ch.26 · brat · nlplabThe automated version of your concept layer (cluster all mentions; link to a canonical entry) and the established way to display coreference chains — but brat is an annotation editor, and standard linking targets a fixed KB, not author-emergent concepts.

H Controlled language & proof rendering

Attempto Controlled English · Univ. Zurich · attempto · Naproche · now in IsabelleA precise English subset parsing to first-order logic; Naproche shows ordinary-looking math prose carrying a checkable structure. Closest in premise — but they require writing in a restricted language, the inverse of overlaying structure on unrestricted prose (the harder, open problem).

Carnap · Fitch-style proof layout · carnap.ioRenders proofs with nested scope lines & per-line feedback — the proof-layout sibling of your nested typed brackets and colored operators.

Caveats the research flagged (kept honest)

Evidence that color-coding actually aids comprehension is thinner than it looks: the one solid HCI study (Sarkar 2015) covers syntax highlighting broadly, not role/bracket coloring; grammar-color research is clinical/anecdotal; Bionic Reading failed large eye-tracking evaluations. → evaluate empirically, don't assume.
Potluck abandoned manual entity-highlighting because it created “hidden state outside the text” — a direct warning for a markup layer like the DSL.
Argument-mapping efficacy (van Gelder ~0.8 SD) is largely the proponents' own evidence.
A few names floated in early searches don't check out: “Legal Begriffsschrift” (no such project), the Fahlman rainbow-parens origin (folklore), Roam's “40% faster” (unverified).

Compiled from three parallel research agents (HCI / augmented-reading · tools-for-thought / notation · argument-mapping / notation), each instructed to prefer primary sources and flag unverifiable claims. Links above are representative, not exhaustive.

← back to the parsed text