Rendering Syntactic Trees with arborize()

The problem

Syntactic trees are a staple of linguistics. Whether you are analyzing argument structure, working through a derivation in a Minimalist framework, or simply illustrating the constituent structure of a sentence for a course handout, you need a way to produce clean, readable tree diagrams.

For a long time, the standard answer was LaTeX. Packages like forest, tikz-qtree, and gb4e have served linguists well, and they remain powerful tools for complex diagrams. But they come with significant overhead: a full LaTeX installation, custom .sty files, and a build pipeline that can be difficult to integrate with modern R-based workflows. If your analysis lives in a Quarto document, a Shiny app, or an R Markdown report, pulling in LaTeX just to draw a tree is a heavy dependency for a relatively simple task.

The situation is further complicated when you want to embed a tree in a format that has nothing to do with LaTeX — a Word document, a Keynote slide, a webpage. A PDF compiled from LaTeX does not travel well.

How arborize() solves it

arborize() takes a syntactic tree description as a character string, renders it using Quarto’s Typst engine, and exports the result as a standalone PNG image. The PNG can be embedded anywhere: in a Quarto document, an R Markdown report, a Word file, a presentation, or uploaded directly to a content management system.

The function handles the entire pipeline internally:

1. It builds a minimal throwaway Quarto document containing the tree
2. Quarto renders the document via Typst to an intermediate PDF
3. The PDF is converted to PNG at the requested resolution
4. Temporary files are cleaned up automatically
5. Optionally, a provenance .yaml file is written alongside the PNG recording the tree string and all rendering arguments

No LaTeX required. No intermediate files left on disk. One function call, one PNG.

Two rendering backends

arborize() supports two rendering backends, controlled by the tree_notation argument.

[S [NP [Det the] [N cat]] [VP [V sat]]]

tree(
  tag: [VP],
  tree(
    tag: [DP],
    [every],
    [farmer]
  ),
  [smiled]
)

Getting the crop right: papersize and margin

Before diving into examples, it is worth understanding how papersize and margin interact to determine what the final PNG looks like.

arborize() renders the tree onto a Typst page of the specified papersize. That page becomes the PDF, which is then converted to PNG. If the page is much larger than the tree, the PNG will have a lot of empty whitespace around the tree — it will look small and lost in the image. If the page is too small, the tree will be clipped.

The goal is to choose a papersize that fits snugly around the tree, with margin providing a comfortable buffer. Think of it as choosing an envelope that is just big enough for your letter.

Examples

arborize(
  "[NP [Det the] [N cat]]",
  output    = "figures/np-tree.png",
  papersize = "a6",
  margin    = "0.5cm"
)

arborize(
  "[S [NP [Det the] [N cat]] [VP [V chased] [NP [Det the] [N dog]]]]",
  output    = "figures/clause-tree.png",
  papersize = "a5",
  margin    = "0.5cm"
)

arborize(
  paste0(
    "[Aspectual Classes ",
    "[Statives [States]] ",
    "[Dynamic ",
    "[Atelic [Activities]] ",
    "[Telic ",
    "[Instantaneous [Achievements]] ",
    "[Durative [Accomplishments]]]]]"
  ),
  output    = "figures/aspectual-classes.png",
  papersize = "a4",
  margin    = "1cm",
  dpi       = 300
)

arborize(
  "[NP [Det the] [N cat]]",
  output    = "figures/np-tree-hires.png",
  papersize = "a6",
  margin    = "0.5cm",
  dpi       = 600
)

arborize(
  "[NP [Det the] [N cat]]",
  output    = "figures/np-tree-tight.png",
  papersize = "a6",
  margin    = "0.1cm"
)

arborize(
  "tree(
    tag: [VP],
    tree(
      tag: [DP],
      [every],
      [farmer]
    ),
    [smiled]
  )",
  tree_notation = "structured",
  output        = "figures/vp-lingotree.png",
  papersize     = "a6",
  margin        = "0.5cm"
)

arborize(
  "[NP [Det the] [N cat]]",
  output     = "figures/np-tree.png",
  papersize  = "a6",
  provenance = FALSE
)

Working with provenance files

When provenance = TRUE (the default), arborize() writes a .yaml file with the same stem as the PNG. For a call like:

arborize(
  "[NP [Det the] [N cat]]",
  output    = "figures/np-tree.png",
  papersize = "a6",
  margin    = "0.5cm"
)

The directory will contain:

figures/
├── np-tree.png
└── np-tree.yaml

The .yaml file records everything needed to reproduce the render:

rendered_by: toolero::arborize(), version 0.4.0
rendered_at: 2026-04-30 14:23:11 CDT
output: /path/to/figures/np-tree.png
tree_notation: simple
typst_package: '@preview/syntree:0.2.1'
dpi: 300
papersize: a6
margin: 0.5cm
tree: '[NP [Det the] [N cat]]'

To re-render from a provenance file — for example to increase the DPI for a journal submission while keeping the exact same tree and crop settings — read the .yaml and pass its fields back to arborize():

p <- yaml::read_yaml("figures/np-tree.yaml")

arborize(
  tree          = p$tree,
  output        = "figures/np-tree-hires.png",
  dpi           = 600,
  tree_notation = p$tree_notation,
  papersize     = p$papersize,
  margin        = p$margin,
  overwrite     = TRUE
)

The provenance file ensures that the high-resolution version is cropped identically to the original — same papersize, same margin, different dpi.

A dedicated rearborize() function that wraps this pattern is not yet implemented. If this would be useful to you, please file an issue at https://github.com/erwinlares/toolero/issues — user demand is the most direct signal that a feature is worth building.

Embedding trees in documents

Once you have a PNG, embedding it in any R-based document is straightforward.

In a Quarto document or R Markdown:

```{{r}}
#| label: fig-np-tree
#| fig-cap: "A simple NP"
#| out-width: "40%"
knitr::include_graphics("figures/np-tree.png")
```

In a Quarto document using Markdown syntax directly:

![A simple NP](figures/np-tree.png){width=40%}

Because arborize() produces a standard PNG, the tree can also be inserted into Word documents via officer, added to a PowerPoint slide via rvg, or uploaded directly to any content management system.

Argument reference

References

• Typst syntree package (v0.2.1): https://typst.app/universe/package/syntree
• Typst lingotree package (v1.0.0): https://typst.app/universe/package/lingotree