Human few-shot learning of compositional instructions

01/14/2019 ∙ by Brenden M. Lake, et al. ∙ 0

People learn in fast and flexible ways that have not been emulated by machines. Once a person learns a new verb "dax," he or she can effortlessly understand how to "dax twice," "walk and dax," or "dax vigorously." There have been striking recent improvements in machine learning for natural language processing, yet the best algorithms require vast amounts of experience and struggle to generalize new concepts in compositional ways. To better understand these distinctively human abilities, we study the compositional skills of people through language-like instruction learning tasks. Our results show that people can learn and use novel functional concepts from very few examples (few-shot learning), successfully applying familiar functions to novel inputs. People can also compose concepts in complex ways that go beyond the provided demonstrations. Two additional experiments examined the assumptions and inductive biases that people make when solving these tasks, revealing three biases: mutual exclusivity, one-to-one mappings, and iconic concatenation. We discuss the implications for cognitive modeling and the potential for building machines with more human-like language learning capabilities.

READ FULL TEXT VIEW PDF
POST COMMENT

Comments

There are no comments yet.

Authors

page 1

page 2

page 3

page 4

This week in AI

Get the week's most popular data science and artificial intelligence research sent straight to your inbox every Saturday.

References

  • Bloom (2000) Bloom:2000Bloom, P.  2000. How Children Learn the Meanings of Words How children learn the meanings of words. Cambridge, MAMIT Press.
  • Calvo  Symons (2014) Calvo:Symons:2014Calvo, P.  Symons, J. ().  2014. The architecture of cognition: Rethinking Fodor and Pylyshyn’s systematicity challenge The architecture of cognition: Rethinking Fodor and Pylyshyn’s systematicity challenge. Cambridge, MAMIT Press.
  • Chomsky (1957) Chomsky:1957Chomsky, N.  1957. Syntactic Structures Syntactic structures. Berlin, GermanyMouton.
  • de Ruiter . (2018) deRuiter:etal:2018de Ruiter, L., Theakston, A., Brandt, S.  Lieven, E.  2018. Iconicity affects children’s comprehension of complex sentences: The role of semantics, clause order, input and individual differences Iconicity affects children’s comprehension of complex sentences: The role of semantics, clause order, input and individual differences. Cognition171202–224.
  • Diesendruck  Markson (2001) Diesendruck2001Diesendruck, G.  Markson, L.  2001. Children’ s Avoidance of Lexical Overlap: A Pragmatic Account Children’ s Avoidance of Lexical Overlap: A Pragmatic Account. Developmental Psychology375630–641.
  • Fitch  Friederici (2012) Fitch:Friederici:2012Fitch, T.  Friederici, A.  2012. Artificial grammar learning meets formal language theory: An overview Artificial grammar learning meets formal language theory: An overview. Philosophical Transactions of the Royal Society B: Biological Sciences36715981933–1955. 10.1098/rstb.2012.0103
  • Fodor (1975) Fodor1975Fodor, J.  1975. The Language of Thought The Language of Thought. Harvard University Press.
  • Fodor  Lepore (2002) Fodor:Lepore:2002Fodor, J.  Lepore, E.  2002. The Compositionality Papers The compositionality papers. Oxford, UKOxford University Press.
  • Fodor  Pylyshyn (1988) Fodor:Pylyshyn:1988Fodor, J.  Pylyshyn, Z.  1988. Connectionism and cognitive architecture: A critical analysis Connectionism and cognitive architecture: A critical analysis. Cognition283–71.
  • Goldin-Meadow . (2008) GoldinMeadow:etal:2008Goldin-Meadow, S., So, WC., Özyürek, A.  Mylander, C.  2008. The natural order of events: How speakers of different languages represent events nonverbally The natural order of events: How speakers of different languages represent events nonverbally. Proceedings of the National Academy of Sciences105279163–9168.
  • Gureckis . (2015) PsiturkGureckis, TM., Martin, J., McDonnell, J., Alexander, RS., Markant, DB., Coenen, A.Chan, P.  2015. psiTurk: An open-source framework for conducting replicable behavioral experiments online psiTurk: An open-source framework for conducting replicable behavioral experiments online. Behavioral Research Methods.
  • Haiman (1980) Haiman:1980Haiman, J.  1980. The iconicity of grammar: Isomorphism and motivation The iconicity of grammar: Isomorphism and motivation. Language563515–540.
  • Lake  Baroni (2018) LakeBaroni2018Lake, BM.  Baroni, M.  2018. Generalization without Systematicity: On the Compositional Skills of Sequence-to-Sequence Recurrent Networks Generalization without Systematicity: On the Compositional Skills of Sequence-to-Sequence Recurrent Networks. International Conference on Machine Learning (ICML). International Conference on Machine Learning (ICML).
  • Lake . (2017) Lake2016Lake, BM., Ullman, TD., Tenenbaum, JB.  Gershman, SJ.  2017. Building machines that learn and think like people Building machines that learn and think like people. Behavioral and Brain Sciences40E253.
  • Loula . (2018) Loula2018Loula, J., Baroni, M.  Lake, BM.  2018. Rearranging the Familiar: Testing Compositional Generalization in Recurrent Networks Rearranging the Familiar: Testing Compositional Generalization in Recurrent Networks. arXiv preprint. http://arxiv.org/abs/1807.07545
  • Macnamara (1982) Macnamara:1982Macnamara, J.  1982. Names for Things: A Study in Human Learning Names for things: A study in human learning. Cambridge, MAMIT Press.
  • Marcus (1998) Marcus1998Marcus, GF.  1998. Rethinking Eliminative Connectionism Rethinking Eliminative Connectionism. Cognitive Psychology28237243–282.
  • Marcus (2003) Marcus2003Marcus, GF.  2003. The Algebraic Mind: Integrating Connectionism and Cognitive Science The Algebraic Mind: Integrating Connectionism and Cognitive Science. Cambridge, MAMIT Press.
  • Marcus . (1999) Marcus1999Marcus, GF., Vijayan, S., Bandi Rao, S.  Vishton, PM.  1999. Rule Learning by Seven-Month-Old Infants Rule Learning by Seven-Month-Old Infants. Science283539877–80.
  • Markman  Wachtel (1988) Markman1988Markman, EM.  Wachtel, GF.  1988. Children’s Use of Mutual Exclusivity to Constrain the Meanings of Words Children’s Use of Mutual Exclusivity to Constrain the Meanings of Words. Cognitive Psychology20121–157.
  • Montague (1970) Montague:1970aMontague, R.  1970. Universal Grammar Universal Grammar. Theoria36373–398.
  • Reber (1967) Reber:1967Reber, A.  1967. Implicit learning of artificial grammars Implicit learning of artificial grammars. Verbal Learning and Verbal Behavior56855–863.
  • Stuhlmuller . (2010) Stuhlmuller2010Stuhlmuller, A., Tenenbaum, JB.  Goodman, ND.  2010. Learning Structured Generative Concepts Learning Structured Generative Concepts. Proceedings of the Thirty-Second Annual Conference of the Cognitive Science Society. Proceedings of the Thirty-Second Annual Conference of the Cognitive Science Society.
  • Sutskever . (2014) Sutskever2014Sutskever, I., Vinyals, O.  Le, QV.  2014. Sequence to Sequence Learning with Neural Networks Sequence to Sequence Learning with Neural Networks. Advances in Neural Information Processing Systems (NIPS). Advances in Neural Information Processing Systems (NIPS).
  • Wonnacott . (2012) Wonnacott:etal:2012Wonnacott, E., Boyd, J., Thomson, J.  Goldberg, A.  2012. Input effects on the acquisition of a novel phrasal construction in 5 year olds Input effects on the acquisition of a novel phrasal construction in 5 year olds. Journal of Memory and Language66458–478.