On the Feasibility of Dependency Parsing of Non-human Sequences Without a Gold Standard

Jul 16, 2026

Dependency parsing aims to recover a dependency tree for a sequence, while unsupervised parsing seeks to learn without a gold-standard treebank. For human languages, evaluation is possible because gold standards usually exist, but for other species such standards are unknown. Using recent network-science advances, the work argues that when non-human primates produce short sequences whose length distributions decay rapidly, parsers can be evaluated by requiring a high proportion of correct edges retrieved. In contrast, human sentences lack this property, making evaluation without gold standards harder.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages40

Reading Time1 h 41 min

Introduction

Dependency parsing and evaluation with/without gold standards

Challenges for other species and low-resource settings

Key argument using sequence-length decay and correct-edge proportion

arXiv :2607 .06542v2 [ cs .CL] 8 Jul 2026 Short paper On the feasibility of dependency parsing of non-human sequences without a gold standard. Is evaluation possible in other species? Ramon Ferrer-i-Cancho∗ , 1, Catherine Hobaiter2, Thore Bergman3 Morgan Gustison4 1 Universitat Politècnica de Catalunya, Department of Computer Science [rferrericancho@cs.upc.edu](rferrericancho@cs.upc.edu) 2 University of St Andrews, School of Psychology and Neuroscience [clh42@st-andrews.ac.uk](clh42@st-andrews.ac.uk) 3 University of Michigan, Departments of Psychology and Ecology and Evolutionary Biology [thore@umich.edu](thore@umich.edu) 4 Western University, Department of Psychology [mgustiso@uwo.ca](mgustiso@uwo.ca) Dependency parsing consists of finding a tree representation for a sequence. Unsupervised dependency parsing aims to develop parsing methods without a gold standard during model training. In human languages, an unsupervised parser can be evaluated because some gold standard is usually available or can be created. For other species, a gold standard is unknown. Thus one may conclude that it is impossible to determine the accuracy of an unsupervised parser and, consequently, dependency parsing is unfeasible in other species. However, here we apply recent advances in network science to demonstrate that the proportion of correct edges retrieved by a parser must be high for the sequences of vocalizations or gestures that non-human primates produce due to the fast decay of the sequence length distribution. In contrast, human language sequences lack that property. Therefore, evaluation without a gold standard is feasible in nonhuman primates but a hard problem in humans. 1. Introduction The syntactic structure of a sentence (Figure 1 (a)) can be represented by a syntactic dependency structure, namely a rooted tree where vertices are words and arcs indicate syntactic dependencies between words as in Figure 1 (b) (Mel’ˇcuk 1988) . A syntactic ∗ Corresponding author Action editor: {action editor name} . Submission received: DD Month YYYY; revised version received: DD Month YYYY; accepted for publication: DD Month YYYY. © 2025 Association for Computational Linguistics Computational Linguistics Volume vv, Number nn (a) Figure 1 (a) The raw sequence of words making the sentence. (b) The syntactic dependency parse of Article 9 of the Universal Declaration of Human Rights by Stanza 1.10.1 [http://stanza.run/](http://stanza.run/). (c) The same parse as in (b) after removing dependency direction. dependency treebank, hereafter a treebank, is a collection of sentences and their dependency structure. Dependency parsing, hereafter parsing, is finding the tree representation of a sentence automatically, and a parse is the tree representation retrieved by a parser for a sentence (Marecek 2012; Carroll 2014). Supervised parsers learn to parse by means of a gold standard, usually a treebank with the correct or desired dependency structure for each sentence (Figure 1 (b)) . Supervised learning consists of fitting a model by means of a treebank as the gold standard. Unsupervised parsers learn to parse without a treebank, namely they learn to parse just from a collection of raw sentences, that is just the word sequence (as in Figure 1 (a)) or the word sequence plus part-of-speech information (Han et al. 2020) . Unsupervised learning consists of fitting a model without a gold standard. Evaluation consists of comparing the parses against a gold standard treebank. Typical evaluation scores are the proportion of trees correctly retrieved or the proportion of dependencies correctly retrieved (Carroll 2014; Han et al. 2020) . In many languages, the evaluation of a parser is straightforward because treebanks are already available. In other languages, evaluation is not possible because treebanks are not forthcoming (the so-called “low-resource” languages), their development is technically costly (Scheinerman 2012) or the lan

FAQs

What problem does the paper address regarding unsupervised dependency parsing?

It addresses how to evaluate an unsupervised dependency parser when no gold-standard dependency structure exists, which is common for non-human species.

Why is evaluation feasible for non-human primates in this work?

Because the sequences produced by non-human primates are short and their sequence-length distributions decay rapidly, which implies that a parser must retrieve a high proportion of correct edges for the observed sequences.

How does the paper compare evaluation feasibility between non-human primates and humans?

It argues that non-human primates satisfy the key property enabling evaluation without a gold standard, while human language sentences lack that property, making evaluation a hard problem.

Jul 16, 2026

Upload by River Wang