A Method for Learning Large-Scale Computational Construction Grammars from Semantically Annotated Corpora

Paul Van Eecke; Katrien Beuls

cs.CLMar 13, 2026

A Method for Learning Large-Scale Computational Construction Grammars from Semantically Annotated Corpora

Authors: Paul Van Eecke, Katrien Beuls

Organizations: Faculté d’informatique, Université de Namur, Rue Grandgagnage 21, B-5000 Namur

Abstract

We present a method for learning large-scale, broad-coverage construction grammars from corpora of language use. Starting from utterances annotated with constituency structure and semantic frames, the method facilitates the learning of human-interpretable computational construction grammars that capture the intricate relationship between syntactic structures and the semantic relations they express. The resulting grammars consist of networks of tens of thousands of constructions formalised within the Fluid Construction Grammar framework. Not only do these grammars support the frame-semantic analysis of open-domain text, they also house a trove of information about the syntactico-semantic usage patterns present in the data they were learnt from. The method and learnt grammars contribute to the scaling of usage-based, constructionist approaches to language, as they corroborate the scalability of a number of fundamental construction grammar conjectures while also providing a practical instrument for the constructionist study of English argument structure in broad-coverage corpora.

Figures & tables

Figure 1: Schematic illustration of the input to the learning process, which consists for each utterance of a syntactic annotation in terms of constituency structure (left) and a semantic annotation in terms PropBank rolesets (right).

Figure 2: Schematic illustration of the frame-evoking tell(verb)-cxn learnt from the input structures shown in Figure 1 . This construction identifies uses of the verb ‘tell’ and indicates their frame-evoking potential.

Figure 3: Schematic illustration of the argument structure construction arg0(np)-v(v)-arg2(np)- arg1(np)-cxn-1 learnt from the structures in Figure 1 . This construction identifies a particular type of ditransitive utterance and labels the constituents that take up the roles of v , arg0 , arg1 and arg2 .

Figure 4: Schematic illustration of the roleset construction tell.01 learnt from the structures in Figure 1 . This construction attributes the tell.01 roleset to roleset instances identified by connected frame-evoking and argument structure constructions.

Figure 5: Schematic illustration of the frame-evoking tell(verb)-cxn , the ditransitive arg0(np)-v(v)-arg2(np)-arg1(np)-cxn-1 and the roleset-attributing tell.01-cxn combining to process the previously unseen utterance “ First, Moses told the people every command in the law. ”. The blue lines connecting the three constructions symbolise their interconnection in the construction network.

Panel of Figure 6 — Figure 6: Frequency of constructions as a function of their rank in the frequency table, revealing their Zipfian distribution. Complete grammar network on log-log scale (left) and 50 most frequent constructions on linear scale (right).

	Precision	Recall	$F_{1}$ score
Roleset	76.15	76.36	76.25
Frame	79.85	80.07	79.96

Table 1: Performance of a ‘pure’ grammar on the task of semantic frame extraction (1,000 held-out utterances).

Semantic Parsing

Semantic parsing maps natural language into structured representations, such as logical forms, graphs, or queries, that machines can execute or reason over.

See all 29 papers

Language Models Learn Constructional Semantics, Not To Mention Syntax: Investigating LM Understanding of Paired-Focus Constructions
May 29, 2026Wesley Scivetti, Ethan Wilcox, Nathan Schneider +2LLM Evaluation Open-Weight Language Models
Georgetown University · The University of Texas at Austin · Leipzig University & ScaDS.AI Dresden/Leipzig
Convergent Representations of Linguistic Constructions in Human and Artificial Neural Systems
Mar 31, 2026Pegah Ramezani, Thomas Kinfe, Andreas Maier +2Neural Encoding Computational Psycholinguistics
Department of English and American Studies, University Erlangen-Nuremberg, Germany · Pattern Recognition Lab, University Erlangen-Nuremberg, Germany · Neuromodulation and Neuroprosthetics, University Hospital Mannheim, University Heidelberg, Germany +2
CxMP: A Linguistic Minimal-Pair Benchmark for Evaluating Constructional Understanding in Language Models
Feb 25, 2026Miyu Oba, Saku SugawaraLLM Evaluation
Nara Institute of Science and Technology · National Institute of Informatics · The University of Tokyo

May 29, 2026J/K move · Enter open · S save

Language Models Learn Constructional Semantics, Not To Mention Syntax: Investigating LM Understanding of Paired-Focus Constructions

Wesley Scivetti, Ethan Wilcox, Nathan Schneider, Kanishka Misra+1

Georgetown University · The University of Texas at Austin · Leipzig University & ScaDS.AI Dresden/Leipzig

LLM Evaluation Open-Weight Language Models

Open paper AskPDF arXiv

Grasping the semantics of rare constructions (form-meaning pairings) has been shown to be a challenging problem that has currently only been solved by the largest LLMs. It remains an open question if open-source models have robust constructional understanding, and if so, what learning dynamics underlie the acquisition of this knowledge. Focusing on a set of rare Paired-Focus constructions in English (e.g. "let alone", "much less"), we construct a novel dataset to test their meanings using both scalar adjectival semantics and general world knowledge. Testing a wide range of models differing in parameter count, architecture, and pretraining dataset size, we find that several modestly sized models are sensitive to both the forms and the meanings of Paired-Focus constructions, though models trained on human-scale data fail at all meaning evaluations. Turning to training dynamics for a set of open-checkpoint models, we find that Paired-Focus understanding emerges later in training than Paired-Focus syntactic knowledge, and that learning of Paired-Focus semantics is correlated with gains in some domains of world knowledge. Overall, our empirical results support the conclusion that modestly sized open-source models can grasp the rare Paired-Focus constructions, and demonstrate a connection between knowledge of Paired-Focus constructions and other meaning domains.

Abstract

Semantic Parsing

Explore similar work

Language Models Learn Constructional Semantics, Not To Mention Syntax: Investigating LM Understanding of Paired-Focus Constructions

Convergent Representations of Linguistic Constructions in Human and Artificial Neural Systems

CxMP: A Linguistic Minimal-Pair Benchmark for Evaluating Constructional Understanding in Language Models