Dimension-Prompts Boost Commonsense Consolidation

Feng, J; Tao, C; Shen, T; Liu, C; Zhao, D

Dimension-Prompts Boost Commonsense Consolidation

Feng, J Tao, C Shen, T Liu, C Zhao, D

Permalink

Publisher:: ACM
Publication Type:: Conference Proceeding
Citation:: SIGIR '23: Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2023, pp. 1934-1938
Issue Date:: 2023-01-01

Closed Access

	Filename	Description	Size
	3539618.3591973.pdf	Published version	957.76 kB	Adobe PDF	View/Open

Copyright Clearance Process

Recently Added
In Progress
Closed Access

This item is closed access and not available.

Full metadata record

Field	Value	Language
dc.contributor.author	Feng, J
dc.contributor.author	Tao, C
dc.contributor.author	Shen, T
dc.contributor.author	Liu, C
dc.contributor.author	Zhao, D
dc.date	2023-07-23
dc.date.accessioned	2024-06-13T03:33:04Z
dc.date.available	2024-06-13T03:33:04Z
dc.date.issued	2023-01-01
dc.identifier.citation	SIGIR '23: Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2023, pp. 1934-1938
dc.identifier.isbn	978-1-4503-9408-6
dc.identifier.uri	http://hdl.handle.net/10453/179510
dc.description.abstract	Neural knowledge models emerged and advanced common-sense-centric knowledge grounding. They parameterize a small seed curated commonsense knowledge graph (CS-KG) in a language model to generalize more. A current trend is to scale the seed up by directly mixing multiple sources of CS-KG (e.g., ATOMIC, ConceptNet) into one model. But, such brute-force mixing inevitably hinders effective knowledge consolidation due to i) ambiguous, polysemic, and/or inconsistent relations across sources and ii) knowledge learned in an entangled manner despite distinct types (e.g., causal, temporal). To mitigate this, we adopt a concept of commonsense knowledge dimension and propose a brand-new dimension-disentangled knowledge model ((DKM)-K-2) learning paradigm with multiple sources. That is, a generative language model with dimension-specific soft prompts is trained to disentangle knowledge acquisitions along with different dimensions and facilitate potential intra-dimension consolidation across CS-KG sources. Experiments show our knowledge model outperforms its baselines in both standard and zero-shot scenarios.
dc.language	en
dc.publisher	ACM
dc.relation.ispartof	SIGIR '23: Proceedings of the 46th International ACM SIGIR Conference on Research and Development in Information Retrieval
dc.relation.ispartof	International ACM SIGIR Conference on Research and Development in Information Retrieval
dc.relation.isbasedon	10.1145/3539618.3591973
dc.rights	info:eu-repo/semantics/closedAccess
dc.title	Dimension-Prompts Boost Commonsense Consolidation
dc.type	Conference Proceeding
utslib.location.activity	Taipei, TAIWAN
pubs.organisational-group	University of Technology Sydney
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology
pubs.organisational-group	University of Technology Sydney/Faculty of Engineering and Information Technology/School of Computer Science
utslib.copyright.status	closed_access	*
pubs.consider-herdc	false
dc.date.updated	2024-06-13T03:33:02Z
pubs.finish-date	2023-07-27
pubs.place-of-publication	USA
pubs.publication-status	Published
pubs.start-date	2023-07-23
dc.location	USA

Abstract:

Neural knowledge models emerged and advanced common-sense-centric knowledge grounding. They parameterize a small seed curated commonsense knowledge graph (CS-KG) in a language model to generalize more. A current trend is to scale the seed up by directly mixing multiple sources of CS-KG (e.g., ATOMIC, ConceptNet) into one model. But, such brute-force mixing inevitably hinders effective knowledge consolidation due to i) ambiguous, polysemic, and/or inconsistent relations across sources and ii) knowledge learned in an entangled manner despite distinct types (e.g., causal, temporal). To mitigate this, we adopt a concept of commonsense knowledge dimension and propose a brand-new dimension-disentangled knowledge model ((DKM)-K-2) learning paradigm with multiple sources. That is, a generative language model with dimension-specific soft prompts is trained to disentangle knowledge acquisitions along with different dimensions and facilitate potential intra-dimension consolidation across CS-KG sources. Experiments show our knowledge model outperforms its baselines in both standard and zero-shot scenarios.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/179510