Verb Argument Construction Project

Overview

We are working on a project to build an inventory of a large number of English verb argument constructions (VACs) using the COBUILD Verb Grammar Patterns descriptions and tools from computational and corpus linguistics. The project has two components: (1) a computational corpus analysis of corpora to retrieve instances of the full range of VACs and (2) psycholinguistic experiments to measure speaker knowledge of these VACs through the verbs selected.

Background

Our experience of language allows us to converge upon similar interpretations of novel utterances like “the ball mandools across the ground” and “the teacher spugged the boy the book.” You know that mandool is a verb of motion and have some idea of how mandooling works – its action semantics. You know that spugging involves some sort of transfer, that the teacher is the donor, the boy the recipient, and that the book is the transferred object. How is this possible, given that you have never heard these verbs before? Each word in these utterances contributes individual meaning, and the verb meanings in these Verb-Argument Constructions (VACs) are usually central. But the larger configuration of words (the construction) has come to carry meaning as a whole (Goldberg, 2003, 2006). The VAC as a category has inherited its schematic meaning from all of the examples you have heard. Mandool inherits its interpretation from the echoes of the verbs that occupy this VAC – words like come, walk, move, …, scud, skitter and flit. Our research explores these processes of usage-based VAC acquisition from large- scale corpus analysis and data from psycholinguistic experiments. In the corpus analyses, we are constructing an inventory from searches of a dependency-parsed version of the British National Corpus (BNC) for specific VACs previously identified in the COBUILD Grammar Patterns volume (Francis, Hunston, & Manning, 1996). Our method applies the following steps.

For each VAC, such as the pattern V(erb) across n(oun phrase):

1. We generate a list of verb types that occupy the construction (e.g. walk, move, skitter).
2. We tally the frequencies of these verbs to produce a frequency ranked type-token profile for these verbs, and we determine the degree to which this is Zipfian (e.g. come 474 … spread 146 … throw 17 … stagger 5) (Zipf, 1949).
3. Because some verbs are faithful to one construction while others are more promiscuous, we produce a contingency-weighted list which reflects their statistical association (e.g. scud, skitter, sprawl, flit have the strongest association with V across n).
4. We use WordNet disambiguated senses of these verbs in the sentences in which they occur to analyze.
5. We use measures of the degree to which there is semantic cohesion of the verbs occupying each construction (e.g., semantic fields TRAVEL and MOVE most frequent for V across n). We measure the degree to which each VAC is more coherent than expected by chance in terms of the association of their grammatical form and semantics, and assess the consequences of this for their acquisition in terms of what is known of the psychology of learning.

In the psycholinguistic experiments we test these predictions and measure the verb-to-construction knowledge of both native and non-native speakers of English. We use free association tasks to direct subjects to think of the first word that comes to mind to fill the V slot in a particular VAC frame. The range of the verbs that they generate, and their speed of access, inform us about the representation

of these VACs in the human mind. We show how the statistical factors in our inventory predict this.

References

Francis, G., Hunston, S., & Manning, E. (Eds.). (1996). Grammar Patterns 1: Verbs. The COBUILD Series. London: Harper Collins.

Goldberg, A. E. (2003). Constructions: a new theoretical approach to language. Trends in Cognitive Science, 7, 219-224.

Goldberg, A. E. (2006). Constructions at work: The nature of generalization in language. Oxford: Oxford University Press.

Zipf, G. K. (1949). Human behaviour and the principle of least effort: An introduction to human ecology. Cambridge, MA: Addison-Wesley.

Project Members

Primary Researchers

• Nick Ellis
• Matt O’Donnell
• Ute Römer

Research Assistants

• Gin Corden (Graduate research assistant)
• Katie Erbach (Undergraduate research assistant)
• Mary Smith (UROP student)
• Danny Tzu-Yu Wu (UM School of Information intern)
• Lucy Zhao (Undergraduate research assistant)

Related presentations

O’Donnell, M.B. & Ellis, N.C. (2010). Towards an Inventory of English Verb Argument Constructions. Paper presented at the Workshop on Extracting and Using Constructions in Computational Linguistics, Human Language Technologies: The 11th Annual Conference of the North American Chapter of the Association for Computational Linguistics, June 2010.

O’Donnell, M.B., Römer, U. & Ellis, N.C. (2011). What do people know about verbs in constructions? Combining corpus and psycholinguistic evidence. Paper to be presented at Alabama symposium on on “Exploring the Boundaries and Applications of Corpus Linguistics”, University of Alabama, AL, April 2011.

O’Donnell, M.B., Römer, U. & Ellis, N.C. (2011). Learning verb-argument constructions: New perspectives from corpus and psycholinguistic analyses. Paper to be presented at ISLE 2 (2nd triennial conference of the International Society for the Linguistics of English), Boston University, MA, June 2011.

O’Donnell, M.B., Römer, U. & Ellis, N.C. (2011). Exploring Zipfian distributions in English verb argument constructions: Corpus and psycholinguistic evidence. Paper presented at Corpus Linguistics 2011, Birmingham, UK, July 2011.

Related publications

Ellis, N.C. & M.B. O’Donnell (Forthcoming). Statistical construction learning: Does a Zipfian problem space ensure robust language learning? In J. Williams & J. Rebuschat (eds.) Statistical Learning and Language Acquisition, Berlin: Mouton de Gruyter.

O’Donnell, M.B. & N.C. Ellis (2010). Towards an Inventory of English Verb Argument Constructions. Proceedings of Workshop on Extracting and Using Constructions in Computational Linguistics NAACL HLT2010: 9-16.