How Efficiency Shapes Human Language

Gibson, Edward; Futrell, Richard; Piantadosi, Steven T.; Dautriche, Isabelle; Mahowald, Kyle; Bergen, Leon; Lévy, Roger

doi:10.1016/j.tics.2019.09.005

Cited by 90 publications

(125 citation statements)

References 0 publications

Supporting

Mentioning

123

Contrasting

Order By: Relevance

“…The rapid timescale of adaptation we have investigated in dyadic reference games is not only of interest in its own right for theories of meaning and social coordination; it is a key building block toward grounding the adaptiveness and efficiency of larger‐scale human language in the cognitive mechanisms of individual minds (Gibson et al, 2019; Kirby et al, 2015). If community‐wide conventions emerge from agents generalizing across different dyadic interactions, then local learning mechanisms leading to efficiency and informativity within a dyad may explain how a community's conventions remain well‐calibrated to the demands of the current environment.…”

Section: Discussionmentioning

confidence: 99%

Characterizing the Dynamics of Learning in Repeated Reference Games

Hawkins¹,

Frank²,

Goodman

2020

Cognitive Science

View full text Add to dashboard Cite

The language we use over the course of conversation changes as we establish common ground and learn what our partner finds meaningful. Here we draw upon recent advances in natural language processing to provide a finer‐grained characterization of the dynamics of this learning process. We release an open corpus (>15,000 utterances) of extended dyadic interactions in a classic repeated reference game task where pairs of participants had to coordinate on how to refer to initially difficult‐to‐describe tangram stimuli. We find that different pairs discover a wide variety of idiosyncratic but efficient and stable solutions to the problem of reference. Furthermore, these conventions are shaped by the communicative context: words that are more discriminative in the initial context (i.e., that are used for one target more than others) are more likely to persist through the final repetition. Finally, we find systematic structure in how a speaker's referring expressions become more efficient over time: Syntactic units drop out in clusters following positive feedback from the listener, eventually leaving short labels containing open‐class parts of speech. These findings provide a higher resolution look at the quantitative dynamics of ad hoc convention formation and support further development of computational models of learning in communication.

show abstract

Section: Discussionmentioning

confidence: 99%

Characterizing the Dynamics of Learning in Repeated Reference Games

Hawkins¹,

Frank²,

Goodman

2020

Cognitive Science

View full text Add to dashboard Cite

show abstract

“…This work, alongside Gildea and Temperley (2010); Liu et al (2017);Futrell et al (2017) helps to validate the extent and pervasiveness of DLM in natural languages. More generally, this literature body has proposed a number of reasons for this behavior, many of which center around the related notions of efficiency (Gibson et al, 2019) and memory constraints (Gulordava et al, 2015) for humans. Recent research at the intersection of psycholinguistics and NLP that has tried to probe for dependencyoriented understanding in neural networks (primarily RNNs) does indicate relationships with specific dependency-types and RNN understanding.…”

Section: Related Workmentioning

confidence: 99%

Long-Distance Dependencies Don’t Have to Be Long: Simplifying through Provably (Approximately) Optimal Permutations

Bommasani¹

2019

Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop

View full text Add to dashboard Cite

Neural models at the sentence level often operate on the constituent words/tokens in a way that encodes the inductive bias of processing the input in a similar fashion to how humans do. However, there is no guarantee that the standard ordering of words is computationally efficient or optimal. To help mitigate this, we consider a dependency parse as a proxy for the inter-word dependencies in a sentence and simplify the sentence with respect to combinatorial objectives imposed on the sentenceparse pair. The associated optimization results in permuted sentences that are provably (approximately) optimal with respect to minimizing dependency parse lengths and that are demonstrably simpler. We evaluate our general-purpose permutations within a finetuning schema for the downstream task of subjectivity analysis. Our fine-tuned baselines reflect a new state of the art for the SUBJ dataset and the permutations we introduce lead to further improvements with a 2.0% increase in classification accuracy (absolute) and a 45% reduction in classification error (relative) over the previous state of the art.

show abstract

“…This approach, termed the Ef f i c i e n c y Hy p o t h e s i s in Futrell (in press), sees languages as shaped by a trade-off between information transfer, ease of production, and ease of comprehension under information processing constraints that are inherent to the human brain. A summary of its basic tenets with extensive supporting references can be found in Gibson et al (2019), see also, e.g., Futrell, Mahowald, and Gibson (2015), Jaeger and Tily (2011), Norcliffe, Harris, and Jaeger (2015), Gibson (1998Gibson ( , 2000, Haspelmath (forthcoming), and Levshina (forthcoming). This approach provides solutions to many traditional problems in psycholinguistics, including the relationship between integration and prediction in processing, between the speaker's needs and the hearer's needs, between ease and complexity of processing and efficient information transfer, and between performance and the grammaticalized conventions of the world's languages.…”

Section: Introductionmentioning

confidence: 99%

English fragments, Minimize Domains, and Minimize Forms

Nykiel

Hawkins

2020

Lang. cogn.

View full text Add to dashboard Cite

We offer an account of preposition drop under clausal ellipsis in terms of two language processing principles: Minimize Domains and Minimize Forms. We argue that when Minimize Domains operates within the PP domain, it disfavors preposition drop due to the preferred independent processability of the PP fragment. When it operates within the VP domain it favors preposition drop in proportion to the number and strength of semantic dependencies between V and P in a given language: the more dependencies there are, and the stronger they are, the stronger the preference for preposition drop. In this way fragments are avoided with long dependencies between P and a distant V. We demonstrate this pattern in English corpora and propose it as an explanation for the typologically unusual preference that English shows for NP fragments. Minimize Forms supports preposition drop in easy-to-process environments cross-linguistically and in English when the more minimal fragment (NP) can be easily linked to its correlate in the antecedent, disfavoring preposition drop elsewhere. The predictions of Minimize Domains and Minimize Forms receive support from a mixed-effects regression model fitted to data from spoken US English, and can be understood as motivations for construction-specific constraints and preferences in clausal ellipsis.

show abstract

How Efficiency Shapes Human Language

Cited by 90 publications

References 0 publications

Characterizing the Dynamics of Learning in Repeated Reference Games

Characterizing the Dynamics of Learning in Repeated Reference Games

Long-Distance Dependencies Don’t Have to Be Long: Simplifying through Provably (Approximately) Optimal Permutations

English fragments, Minimize Domains, and Minimize Forms

Contact Info

Product

Resources

About