Improving zero-shot learning baselines with commonsense knowledge
Zero-shot learning — the problem of training and testing on a completely disjoint set of classes — relies greatly on its ability to transfer knowledge from train classes to test classes. Traditionally semantic embeddings consisting of human-defined attributes or distributed word embeddings are used...
Saved in:
Main Authors: | , , , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2023
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/170538 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-170538 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1705382023-09-19T01:58:13Z Improving zero-shot learning baselines with commonsense knowledge Roy, Abhinaba Ghosal, Deepanway Cambria, Erik Majumder, Navonil Mihalcea, Rada Poria, Soujanya School of Computer Science and Engineering Engineering::Computer science and engineering Commonsense Knowledge Zero-shot Learning Zero-shot learning — the problem of training and testing on a completely disjoint set of classes — relies greatly on its ability to transfer knowledge from train classes to test classes. Traditionally semantic embeddings consisting of human-defined attributes or distributed word embeddings are used to facilitate this transfer by improving the association between visual and semantic embeddings. In this paper, we take advantage of explicit relations between nodes defined in ConceptNet, a commonsense knowledge graph, to generate commonsense embeddings of the class labels by using a graph convolution network-based autoencoder. Our experiments performed on three standard benchmark datasets surpass the strong baselines when we fuse our commonsense embeddings with existing semantic embeddings, i.e., human-defined attributes and distributed word embeddings. This work paves the path to more brain-inspired approaches to zero-short learning. Agency for Science, Technology and Research (A*STAR) Ministry of Education (MOE) This research is supported by the Agency for Science, Technology and Research (A*STAR) under its AME Programmatic Funding Scheme (Project #A18A2b0046) and Project T2MOE2008 awarded by Singapore's MoE under its Tier-2 Grant Scheme. 2023-09-19T01:58:13Z 2023-09-19T01:58:13Z 2022 Journal Article Roy, A., Ghosal, D., Cambria, E., Majumder, N., Mihalcea, R. & Poria, S. (2022). Improving zero-shot learning baselines with commonsense knowledge. Cognitive Computation, 14(6), 2212-2222. https://dx.doi.org/10.1007/s12559-022-10044-0 1866-9956 https://hdl.handle.net/10356/170538 10.1007/s12559-022-10044-0 2-s2.0-85134593579 6 14 2212 2222 en A18A2b0046 T2MOE2008 Cognitive Computation © 2022 The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
continent |
Asia |
country |
Singapore Singapore |
content_provider |
NTU Library |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Computer science and engineering Commonsense Knowledge Zero-shot Learning |
spellingShingle |
Engineering::Computer science and engineering Commonsense Knowledge Zero-shot Learning Roy, Abhinaba Ghosal, Deepanway Cambria, Erik Majumder, Navonil Mihalcea, Rada Poria, Soujanya Improving zero-shot learning baselines with commonsense knowledge |
description |
Zero-shot learning — the problem of training and testing on a completely disjoint set of classes — relies greatly on its ability to transfer knowledge from train classes to test classes. Traditionally semantic embeddings consisting of human-defined attributes or distributed word embeddings are used to facilitate this transfer by improving the association between visual and semantic embeddings. In this paper, we take advantage of explicit relations between nodes defined in ConceptNet, a commonsense knowledge graph, to generate commonsense embeddings of the class labels by using a graph convolution network-based autoencoder. Our experiments performed on three standard benchmark datasets surpass the strong baselines when we fuse our commonsense embeddings with existing semantic embeddings, i.e., human-defined attributes and distributed word embeddings. This work paves the path to more brain-inspired approaches to zero-short learning. |
author2 |
School of Computer Science and Engineering |
author_facet |
School of Computer Science and Engineering Roy, Abhinaba Ghosal, Deepanway Cambria, Erik Majumder, Navonil Mihalcea, Rada Poria, Soujanya |
format |
Article |
author |
Roy, Abhinaba Ghosal, Deepanway Cambria, Erik Majumder, Navonil Mihalcea, Rada Poria, Soujanya |
author_sort |
Roy, Abhinaba |
title |
Improving zero-shot learning baselines with commonsense knowledge |
title_short |
Improving zero-shot learning baselines with commonsense knowledge |
title_full |
Improving zero-shot learning baselines with commonsense knowledge |
title_fullStr |
Improving zero-shot learning baselines with commonsense knowledge |
title_full_unstemmed |
Improving zero-shot learning baselines with commonsense knowledge |
title_sort |
improving zero-shot learning baselines with commonsense knowledge |
publishDate |
2023 |
url |
https://hdl.handle.net/10356/170538 |
_version_ |
1779156304070180864 |