Retrodiction as delayed recurrence: the case of adjectives in italian and english
R. Alhama, F. Zermiani, and A. Khaliq. Proceedings of the The 19th Annual Workshop of the Australasian Language Technology Association, page 163--168. Australasian Language Technology Association, (2021)
Abstract
We address the question of how to account for both forward and backward dependencies in an online processing account of human language acquisition. We focus on descriptive adjectives in English and Italian, and show that the acquisition of adjectives in these languages likely relies on tracking both forward and backward regularities. Our simulations confirm that forward-predicting models like standard Recurrent Neural Networks (RNN) cannot account for this phenomenon due to the lack of backward prediction, but the addition of a small delay (as proposed in Turek et al., 2019) endows the RNN with the ability to not only predict but also retrodict.
%0 Conference Paper
%1 alhama2021retrodiction
%A Alhama, Raquel G
%A Zermiani, Francesca
%A Khaliq, Atiqah
%B Proceedings of the The 19th Annual Workshop of the Australasian Language Technology Association
%D 2021
%I Australasian Language Technology Association
%K childdirectedspeech nlp retrodiction
%P 163--168
%T Retrodiction as delayed recurrence: the case of adjectives in italian and english
%U https://aclanthology.org/2021.alta-1.17
%X We address the question of how to account for both forward and backward dependencies in an online processing account of human language acquisition. We focus on descriptive adjectives in English and Italian, and show that the acquisition of adjectives in these languages likely relies on tracking both forward and backward regularities. Our simulations confirm that forward-predicting models like standard Recurrent Neural Networks (RNN) cannot account for this phenomenon due to the lack of backward prediction, but the addition of a small delay (as proposed in Turek et al., 2019) endows the RNN with the ability to not only predict but also retrodict.
@inproceedings{alhama2021retrodiction,
abstract = {We address the question of how to account for both forward and backward dependencies in an online processing account of human language acquisition. We focus on descriptive adjectives in English and Italian, and show that the acquisition of adjectives in these languages likely relies on tracking both forward and backward regularities. Our simulations confirm that forward-predicting models like standard Recurrent Neural Networks (RNN) cannot account for this phenomenon due to the lack of backward prediction, but the addition of a small delay (as proposed in Turek et al., 2019) endows the RNN with the ability to not only predict but also retrodict.},
added-at = {2024-03-15T22:58:33.000+0100},
author = {Alhama, Raquel G and Zermiani, Francesca and Khaliq, Atiqah},
biburl = {https://puma.ub.uni-stuttgart.de/bibtex/22f6d6ee52efcb0247d36072e7f82ac3a/fzermiani},
booktitle = {Proceedings of the The 19th Annual Workshop of the Australasian Language Technology Association},
description = {Visualizza articolo},
interhash = {59e6911b97643670e34d437383e721d1},
intrahash = {2f6d6ee52efcb0247d36072e7f82ac3a},
keywords = {childdirectedspeech nlp retrodiction},
pages = {163--168},
publisher = {Australasian Language Technology Association},
timestamp = {2024-03-15T22:58:33.000+0100},
title = {Retrodiction as delayed recurrence: the case of adjectives in italian and english},
url = {https://aclanthology.org/2021.alta-1.17},
year = 2021
}