@inproceedings{lee-etal-2021-learning,
title = "Learning to Select Question-Relevant Relations for Visual Question Answering",
author = "Lee, Jaewoong and
Lee, Heejoon and
Lee, Hwanhee and
Jung, Kyomin",
editor = "Zadeh, Amir and
Morency, Louis-Philippe and
Liang, Paul Pu and
Ross, Candace and
Salakhutdinov, Ruslan and
Poria, Soujanya and
Cambria, Erik and
Shi, Kelly",
booktitle = "Proceedings of the Third Workshop on Multimodal Artificial Intelligence",
month = jun,
year = "2021",
address = "Mexico City, Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2021.maiworkshop-1.13/",
doi = "10.18653/v1/2021.maiworkshop-1.13",
pages = "87--96",
abstract = "Previous existing visual question answering (VQA) systems commonly use graph neural networks(GNNs) to extract visual relationships such as semantic relations or spatial relations. However, studies that use GNNs typically ignore the importance of each relation and simply concatenate outputs from multiple relation encoders. In this paper, we propose a novel layer architecture that fuses multiple visual relations through an attention mechanism to address this issue. Specifically, we develop a model that uses question embedding and joint embedding of the encoders to obtain dynamic attention weights with regard to the type of questions. Using the learnable attention weights, the proposed model can efficiently use the necessary visual relation features for a given question. Experimental results on the VQA 2.0 dataset demonstrate that the proposed model outperforms existing graph attention network-based architectures. Additionally, we visualize the attention weight and show that the proposed model assigns a higher weight to relations that are more relevant to the question."
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lee-etal-2021-learning">
<titleInfo>
<title>Learning to Select Question-Relevant Relations for Visual Question Answering</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jaewoong</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heejoon</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hwanhee</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kyomin</namePart>
<namePart type="family">Jung</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2021-06</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Third Workshop on Multimodal Artificial Intelligence</title>
</titleInfo>
<name type="personal">
<namePart type="given">Amir</namePart>
<namePart type="family">Zadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Louis-Philippe</namePart>
<namePart type="family">Morency</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Paul</namePart>
<namePart type="given">Pu</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Candace</namePart>
<namePart type="family">Ross</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruslan</namePart>
<namePart type="family">Salakhutdinov</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Soujanya</namePart>
<namePart type="family">Poria</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Erik</namePart>
<namePart type="family">Cambria</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kelly</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Mexico City, Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>Previous existing visual question answering (VQA) systems commonly use graph neural networks(GNNs) to extract visual relationships such as semantic relations or spatial relations. However, studies that use GNNs typically ignore the importance of each relation and simply concatenate outputs from multiple relation encoders. In this paper, we propose a novel layer architecture that fuses multiple visual relations through an attention mechanism to address this issue. Specifically, we develop a model that uses question embedding and joint embedding of the encoders to obtain dynamic attention weights with regard to the type of questions. Using the learnable attention weights, the proposed model can efficiently use the necessary visual relation features for a given question. Experimental results on the VQA 2.0 dataset demonstrate that the proposed model outperforms existing graph attention network-based architectures. Additionally, we visualize the attention weight and show that the proposed model assigns a higher weight to relations that are more relevant to the question.</abstract>
<identifier type="citekey">lee-etal-2021-learning</identifier>
<identifier type="doi">10.18653/v1/2021.maiworkshop-1.13</identifier>
<location>
<url>https://aclanthology.org/2021.maiworkshop-1.13/</url>
</location>
<part>
<date>2021-06</date>
<extent unit="page">
<start>87</start>
<end>96</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Learning to Select Question-Relevant Relations for Visual Question Answering
%A Lee, Jaewoong
%A Lee, Heejoon
%A Lee, Hwanhee
%A Jung, Kyomin
%Y Zadeh, Amir
%Y Morency, Louis-Philippe
%Y Liang, Paul Pu
%Y Ross, Candace
%Y Salakhutdinov, Ruslan
%Y Poria, Soujanya
%Y Cambria, Erik
%Y Shi, Kelly
%S Proceedings of the Third Workshop on Multimodal Artificial Intelligence
%D 2021
%8 June
%I Association for Computational Linguistics
%C Mexico City, Mexico
%F lee-etal-2021-learning
%X Previous existing visual question answering (VQA) systems commonly use graph neural networks(GNNs) to extract visual relationships such as semantic relations or spatial relations. However, studies that use GNNs typically ignore the importance of each relation and simply concatenate outputs from multiple relation encoders. In this paper, we propose a novel layer architecture that fuses multiple visual relations through an attention mechanism to address this issue. Specifically, we develop a model that uses question embedding and joint embedding of the encoders to obtain dynamic attention weights with regard to the type of questions. Using the learnable attention weights, the proposed model can efficiently use the necessary visual relation features for a given question. Experimental results on the VQA 2.0 dataset demonstrate that the proposed model outperforms existing graph attention network-based architectures. Additionally, we visualize the attention weight and show that the proposed model assigns a higher weight to relations that are more relevant to the question.
%R 10.18653/v1/2021.maiworkshop-1.13
%U https://aclanthology.org/2021.maiworkshop-1.13/
%U https://doi.org/10.18653/v1/2021.maiworkshop-1.13
%P 87-96
Markdown (Informal)
[Learning to Select Question-Relevant Relations for Visual Question Answering](https://aclanthology.org/2021.maiworkshop-1.13/) (Lee et al., maiworkshop 2021)
ACL