Ontological Knowledge Learning for Relation Extraction

Li, Yang

Ontological Knowledge Learning for Relation Extraction

Li, Yang

Permalink

Publication Type:: Thesis
Issue Date:: 2023

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (3.01 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Li, Yang
dc.date.accessioned	2024-04-12T03:22:54Z
dc.date.available	2024-04-12T03:22:54Z
dc.date.issued	2023
dc.identifier.uri	http://hdl.handle.net/10453/177813
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	The abundance of unstructured text online necessitates efficient Information Extraction (IE) methods. Relation Extraction (RE), a crucial IE sub-task, identifies relationships between entities, benefiting numerous Natural Language Processing (NLP) applications. RE, discerning entity relations in text, faces the demand of extensive training data. To address this, Distant Supervision (DS) is proposed but it leads to the wrong labelling problem and the long-tail relations. Previous works have employed Multi-Instance Learning (MIL) and selective attention networks for reducing noisy labelling, with extra ontological knowledge from Knowledge Graphs (KGs) enhancing entity pair understanding and leveraging relational hierarchies. We present progressive advancements in RE. It introduces a novel lightweight framework to better utilize ontological knowledge in texts. Building upon this, the research further innovates with a relation-augmented attention mechanism with effective hierarchical relation sharing for addressing the long-tail relations. Furthermore, the thesis employs specific prefix tuning with Pre-trained Language Models (PLMs) in limited data scenarios. This approach harnesses ontological knowledge, focusing on contrasting attributes between factual and counterfactual scenarios, significantly enhancing RE task performance.	en_US.UTF-8
dc.format	Thesis (PhD)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/177813/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2023 Yang Li
dc.rights	au.edu.uts.lib/cph
dc.title	Ontological Knowledge Learning for Relation Extraction	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

The abundance of unstructured text online necessitates efficient Information Extraction (IE) methods. Relation Extraction (RE), a crucial IE sub-task, identifies relationships between entities, benefiting numerous Natural Language Processing (NLP) applications. RE, discerning entity relations in text, faces the demand of extensive training data. To address this, Distant Supervision (DS) is proposed but it leads to the wrong labelling problem and the long-tail relations. Previous works have employed Multi-Instance Learning (MIL) and selective attention networks for reducing noisy labelling, with extra ontological knowledge from Knowledge Graphs (KGs) enhancing entity pair understanding and leveraging relational hierarchies. We present progressive advancements in RE. It introduces a novel lightweight framework to better utilize ontological knowledge in texts. Building upon this, the research further innovates with a relation-augmented attention mechanism with effective hierarchical relation sharing for addressing the long-tail relations. Furthermore, the thesis employs specific prefix tuning with Pre-trained Language Models (PLMs) in limited data scenarios. This approach harnesses ontological knowledge, focusing on contrasting attributes between factual and counterfactual scenarios, significantly enhancing RE task performance.

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

http://hdl.handle.net/10453/177813