TY - GEN
T1 - Rearranging the Familiar
T2 - 1st Workshop on BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, co-located with the 2018 Conference on Empirical Methods in Natural Language Processing, EMNLP 2018
AU - Loula, João
AU - Baroni, Marco
AU - Lake, Brenden M.
N1 - Publisher Copyright:
© 2018 Association for Computational Linguistics
PY - 2018
Y1 - 2018
N2 - Systematic compositionality is the ability to recombine meaningful units with regular and predictable outcomes, and it's seen as key to the human capacity for generalization in language. Recent work (Lake and Baroni, 2018) has studied systematic compositionality in modern seq2seq models using generalization to novel navigation instructions in a grounded environment as a probing tool. Lake and Baroni's main experiment required the models to quickly bootstrap the meaning of new words. We extend this framework here to settings where the model needs only to recombine well-trained functional words (such as “around” and “right”) in novel contexts. Our findings confirm and strengthen the earlier ones: seq2seq models can be impressively good at generalizing to novel combinations of previously-seen input, but only when they receive extensive training on the specific pattern to be generalized (e.g., generalizing from many examples of “X around right” to “jump around right”), while failing when generalization requires novel application of compositional rules (e.g., inferring the meaning of “around right” from those of “right” and “around”).
AB - Systematic compositionality is the ability to recombine meaningful units with regular and predictable outcomes, and it's seen as key to the human capacity for generalization in language. Recent work (Lake and Baroni, 2018) has studied systematic compositionality in modern seq2seq models using generalization to novel navigation instructions in a grounded environment as a probing tool. Lake and Baroni's main experiment required the models to quickly bootstrap the meaning of new words. We extend this framework here to settings where the model needs only to recombine well-trained functional words (such as “around” and “right”) in novel contexts. Our findings confirm and strengthen the earlier ones: seq2seq models can be impressively good at generalizing to novel combinations of previously-seen input, but only when they receive extensive training on the specific pattern to be generalized (e.g., generalizing from many examples of “X around right” to “jump around right”), while failing when generalization requires novel application of compositional rules (e.g., inferring the meaning of “around right” from those of “right” and “around”).
UR - http://www.scopus.com/inward/record.url?scp=85120741456&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85120741456&partnerID=8YFLogxK
M3 - Conference contribution
AN - SCOPUS:85120741456
T3 - EMNLP 2018 - 2018 EMNLP Workshop BlackboxNLP: Analyzing and Interpreting Neural Networks for NLP, Proceedings of the 1st Workshop
SP - 108
EP - 114
BT - EMNLP 2018 - 2018 EMNLP Workshop BlackboxNLP
PB - Association for Computational Linguistics (ACL)
Y2 - 1 November 2018
ER -