@inproceedings{bhatt-etal-2020-much,
title = "How much complexity does an {RNN} architecture need to learn syntax-sensitive dependencies?",
author = "Bhatt, Gantavya and
Bansal, Hritik and
Singh, Rishubh and
Agarwal, Sumeet",
editor = "Rijhwani, Shruti and
Liu, Jiangming and
Wang, Yizhong and
Dror, Rotem",
booktitle = "Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop",
month = jul,
year = "2020",
address = "Online",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2020.acl-srw.33",
doi = "10.18653/v1/2020.acl-srw.33",
pages = "244--254",
abstract = "Long short-term memory (LSTM) networks and their variants are capable of encapsulating long-range dependencies, which is evident from their performance on a variety of linguistic tasks. On the other hand, simple recurrent networks (SRNs), which appear more biologically grounded in terms of synaptic connections, have generally been less successful at capturing long-range dependencies as well as the loci of grammatical errors in an unsupervised setting. In this paper, we seek to develop models that bridge the gap between biological plausibility and linguistic competence. We propose a new architecture, the Decay RNN, which incorporates the decaying nature of neuronal activations and models the excitatory and inhibitory connections in a population of neurons. Besides its biological inspiration, our model also shows competitive performance relative to LSTMs on subject-verb agreement, sentence grammaticality, and language modeling tasks. These results provide some pointers towards probing the nature of the inductive biases required for RNN architectures to model linguistic phenomena successfully.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="bhatt-etal-2020-much">
<titleInfo>
<title>How much complexity does an RNN architecture need to learn syntax-sensitive dependencies?</title>
</titleInfo>
<name type="personal">
<namePart type="given">Gantavya</namePart>
<namePart type="family">Bhatt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hritik</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rishubh</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sumeet</namePart>
<namePart type="family">Agarwal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2020-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shruti</namePart>
<namePart type="family">Rijhwani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiangming</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yizhong</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rotem</namePart>
<namePart type="family">Dror</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Online</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Long short-term memory (LSTM) networks and their variants are capable of encapsulating long-range dependencies, which is evident from their performance on a variety of linguistic tasks. On the other hand, simple recurrent networks (SRNs), which appear more biologically grounded in terms of synaptic connections, have generally been less successful at capturing long-range dependencies as well as the loci of grammatical errors in an unsupervised setting. In this paper, we seek to develop models that bridge the gap between biological plausibility and linguistic competence. We propose a new architecture, the Decay RNN, which incorporates the decaying nature of neuronal activations and models the excitatory and inhibitory connections in a population of neurons. Besides its biological inspiration, our model also shows competitive performance relative to LSTMs on subject-verb agreement, sentence grammaticality, and language modeling tasks. These results provide some pointers towards probing the nature of the inductive biases required for RNN architectures to model linguistic phenomena successfully.</abstract>
<identifier type="citekey">bhatt-etal-2020-much</identifier>
<identifier type="doi">10.18653/v1/2020.acl-srw.33</identifier>
<location>
<url>https://aclanthology.org/2020.acl-srw.33</url>
</location>
<part>
<date>2020-07</date>
<extent unit="page">
<start>244</start>
<end>254</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T How much complexity does an RNN architecture need to learn syntax-sensitive dependencies?
%A Bhatt, Gantavya
%A Bansal, Hritik
%A Singh, Rishubh
%A Agarwal, Sumeet
%Y Rijhwani, Shruti
%Y Liu, Jiangming
%Y Wang, Yizhong
%Y Dror, Rotem
%S Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics: Student Research Workshop
%D 2020
%8 July
%I Association for Computational Linguistics
%C Online
%F bhatt-etal-2020-much
%X Long short-term memory (LSTM) networks and their variants are capable of encapsulating long-range dependencies, which is evident from their performance on a variety of linguistic tasks. On the other hand, simple recurrent networks (SRNs), which appear more biologically grounded in terms of synaptic connections, have generally been less successful at capturing long-range dependencies as well as the loci of grammatical errors in an unsupervised setting. In this paper, we seek to develop models that bridge the gap between biological plausibility and linguistic competence. We propose a new architecture, the Decay RNN, which incorporates the decaying nature of neuronal activations and models the excitatory and inhibitory connections in a population of neurons. Besides its biological inspiration, our model also shows competitive performance relative to LSTMs on subject-verb agreement, sentence grammaticality, and language modeling tasks. These results provide some pointers towards probing the nature of the inductive biases required for RNN architectures to model linguistic phenomena successfully.
%R 10.18653/v1/2020.acl-srw.33
%U https://aclanthology.org/2020.acl-srw.33
%U https://doi.org/10.18653/v1/2020.acl-srw.33
%P 244-254
Markdown (Informal)
[How much complexity does an RNN architecture need to learn syntax-sensitive dependencies?](https://aclanthology.org/2020.acl-srw.33) (Bhatt et al., ACL 2020)
ACL