Can it address leaves as well as internal nodes?

Yes. Any node reached by a valid sequence of child choices has an address.

Is the root always written 0?

No. Conventions differ; this entry uses ε for the empty path and starts child indices at zero.

Can two nodes have the same address in one tree?

Under a fixed convention, each valid path selects one node. Repeated labels do not change that uniqueness.

AI Glossary / Natural language and speech / Intermediate

Gorn address

Definition

A Gorn address identifies a node in an ordered rooted tree by recording the sequence of child choices needed to reach it from the root. It is useful for referring precisely to locations in parse trees and other tree structures. The root notation and child-numbering convention must be stated because published conventions differ.

What it means

A tree can contain several nodes with the same label, such as multiple noun phrases. A label alone therefore may not identify the intended node. A path address distinguishes the nodes by structural position.
In the convention used throughout this entry, the root has the empty address, written ε, and children are numbered from zero in left-to-right order. The first child has address 0, the second 1, and so on. Dots separate successive choices.
An address is a sequence, not a decimal number. Thus 1.0 means take the second child and then its first child. It does not represent the numeric value one.

How it works

Choose an ordered rooted tree and a numbering convention. Assign the empty sequence to the root. For every child, append its child index to the parent's sequence.
To locate a node, begin at the root and follow the indices one at a time. If an indicated child does not exist, the address is invalid for that tree.
Removing the final index gives a non-root node's parent address. A node's address is a prefix of the addresses of its descendants, which makes some structural relationships easy to express.
An address is meaningful relative to a particular tree and version. Reordering or inserting siblings can change the address of an otherwise recognizable node, so path addresses should not automatically be treated as permanent identifiers.

Three practical examples

A parse-tree annotator points to one of several noun-phrase nodes without relying on its repeated label.
A tree-transformation rule identifies the exact subtree that should be replaced or inspected.
A debugging report records the path to a malformed node so another developer can reproduce the structural check on the same tree.

Think of it this way

Think of directions through a building: take the second corridor, then the first doorway. The directions identify a route through an ordered structure. The analogy is limited because a Gorn address follows a tree's unique root-to-node path, while real buildings can contain loops, multiple entrances, and several routes to the same room.

Worked example

Build a small ordered tree with root S. Its first child is NP and its second child is VP. NP has two children, Det and N. VP also has two children, V and another NP. That final NP has one child, N.
Using our zero-based convention, the root is ε. The left NP is 0 and VP is 1. Det is 0.0, the first N is 0.1, V is 1.0, the right NP is 1.1, and its N is 1.1.0.
To find 1.1.0, start at S, choose its second child VP, choose VP's second child NP, and choose that NP's first child N. Three indices mean the node is three edges below the root.
The parent of 1.1.0 is 1.1. The node at 1 is an ancestor because its address is a proper prefix. Address 1.2 is invalid in this tree because VP has only children 0 and 1.
If a new child is inserted before VP under S, VP becomes child 2 and the former 1.1.0 node moves to 2.1.0. Its label can remain N while its path address changes.

Illustrative case study

In a fictional project, Ash Syntax reviews automatically produced parse trees. Review comments initially say only “check the NP,” which is ambiguous when a sentence contains several noun phrases.
The team adds a tree identifier, version, and Gorn address to each comment. After a structural edit, the software marks older path references for reconciliation instead of silently assuming that they still identify the same node.
The imagined workflow demonstrates precise location reporting and version awareness. It does not claim that path addresses validate the linguistic analysis or that a node at a valid address necessarily has the correct label.

Common misconceptions and mistakes

A Gorn address is not a globally unique identifier unless additional information identifies the tree and convention.
The digits do not encode the node's grammatical meaning. The same address can point to different labels in different trees.
Zero-based and one-based schemes should not be mixed. Some presentations also use a special root label such as 0, which requires different handling from the empty-root convention used here.
Dots delimit path components; they are not decimal points. Explicit separators are particularly helpful when a node has ten or more children.
Ordinary Gorn addressing assumes a tree. A shared-node graph can have multiple root-to-node paths and requires a clearly defined addressing policy.

Business use

This is a technical notation with practical value in annotation, parser maintenance, and tree-processing tools. It can make an error report or edit request reproducible when several structural nodes look alike.
For an annotation platform, document the root and indexing conventions and preserve the tree version alongside an address. If users need references that survive edits, combine path locations with a separate identity or migration mechanism rather than assuming positional references are stable.

Check your understanding

Under this entry's convention, what is the parent of the node at 2.0.3, and how deep is that node?

Answer: Its parent is 2.0. The node is three edges below the root because the address contains three child choices.

Related terms

Parse tree
rooted tree
ordered tree
subtree
tree traversal
tree-adjoining grammar
node identifier
prefix relation

Learn next

draw a tree with repeated labels and assign addresses to every node. Then insert a sibling and track which addresses change, distinguishing structural position from persistent identity.

Sources

Graf, “Subregular Tree Transductions, Movement, Copies, Traces, and the Ban on Improper Movement,” original research definition of Gorn addresses: https://aclanthology.org/2023.scil-1.25.pdf. University of Freiburg, “Lexicalized Tree-Adjoining Grammars,” original teaching slides showing a different root convention: https://ad-teaching.cs.uni-freiburg.de/efficient-nlp-ws1112/session-10.pdf. The tree example above is original. Independent source review: Graf’s Gorn-address definition and Freiburg’s alternative root convention were read; this entry consistently uses an empty root and zero-based child indices.

Previous: GOP (group of pictures) Next: Gossip protocol