English
 
Help Privacy Policy Disclaimer
  Advanced SearchBrowse

Item

ITEM ACTIONSEXPORT

Released

Conference Paper

Structure-(in)dependent interpretation of phrases in humans and LSTMs

MPS-Authors
/persons/resource/persons225918

Coopmans,  Cas W.
Neurobiology of Language Department, MPI for Psycholinguistics, Max Planck Society;
Centre for Language Studies, External Organizations;
Language and Computation in Neural Systems, MPI for Psycholinguistics, Max Planck Society;
International Max Planck Research School for Language Sciences, MPI for Psycholinguistics, Max Planck Society;

Kaushik,  Karthikeya
Language and Computation in Neural Systems, MPI for Psycholinguistics, Max Planck Society;

/persons/resource/persons69

Hagoort,  Peter
Neurobiology of Language Department, MPI for Psycholinguistics, Max Planck Society;
Donders Institute for Brain, Cognition and Behaviour, External Organizations;

/persons/resource/persons198520

Martin,  Andrea E.
Language and Computation in Neural Systems, MPI for Psycholinguistics, Max Planck Society;
Donders Institute for Brain, Cognition and Behaviour, External Organizations;

External Resource

full text via ScholarWorks@UMass Amherst
(Supplementary material)

Supplementary Material (public)
There is no public supplementary material available
Citation

Coopmans, C. W., De Hoop, H., Kaushik, K., Hagoort, P., & Martin, A. E. (2021). Structure-(in)dependent interpretation of phrases in humans and LSTMs. In Proceedings of the Society for Computation in Linguistics (SCiL 2021) (pp. 459-463).


Cite as: http://hdl.handle.net/21.11116/0000-0008-139C-3
Abstract
In this study, we compared the performance of a long short-term memory (LSTM) neural network to the behavior of human participants on a language task that requires hierarchically structured knowledge. We show that humans interpret ambiguous noun phrases, such as second blue ball, in line with their hierarchical constituent structure. LSTMs, instead, only do so after unambiguous training, and they do not systematically generalize to novel items. Overall, the results of our simulations indicate that a model can behave hierarchically without relying on hierarchical constituent structure.