Meta-based self-training and re-weighting for aspect-based sentiment analysis

Aspect-based sentiment analysis (ABSA) means to identify fine-grained aspects, opinions, and sentiment polarities. Recent ABSA research focuses on utilizing multi-task learning (MTL) to achieve less computational costs and better performance. However, there are certain limits in MTL-based ABSA. For...

Popoln opis

Bibliografske podrobnosti
Main Authors: He, Kai, Mao, Rui, Gong, Tieliang, Li, Chen, Cambria, Erik
Drugi avtorji: School of Computer Science and Engineering
Format: Journal Article
Jezik:English
Izdano: 2022
Teme:
Online dostop:https://hdl.handle.net/10356/163145
_version_ 1826121746051235840
author He, Kai
Mao, Rui
Gong, Tieliang
Li, Chen
Cambria, Erik
author2 School of Computer Science and Engineering
author_facet School of Computer Science and Engineering
He, Kai
Mao, Rui
Gong, Tieliang
Li, Chen
Cambria, Erik
author_sort He, Kai
collection NTU
description Aspect-based sentiment analysis (ABSA) means to identify fine-grained aspects, opinions, and sentiment polarities. Recent ABSA research focuses on utilizing multi-task learning (MTL) to achieve less computational costs and better performance. However, there are certain limits in MTL-based ABSA. For example, unbalanced labels and sub-task learning difficulties may result in the biases that some labels and sub-tasks are overfitting, while the others are underfitting. To address these issues, inspired by neuro-symbolic learning systems, we propose a meta-based self-training method with a meta-weighter (MSM). We believe that a generalizable model can be achieved by appropriate symbolic representation selection (in-domain knowledge) and effective learning control (regulation) in a neural system. Thus, MSM trains a teacher model to generate in-domain knowledge (e.g., unlabeled data selection and pseudo-label generation), where the generated pseudo-labels are used by a student model for supervised learning. Then, the meta-weighter of MSM is jointly trained with the student model to provide each instance with sub-task-specific weights to coordinate their convergence rates, balancing class labels, and alleviating noise impacts introduced from self-training. The following experiments indicate that MSM can utilize 50% labeled data to achieve comparable results to state-of-arts models in ABSA and outperform them with all labeled data.
first_indexed 2024-10-01T05:37:12Z
format Journal Article
id ntu-10356/163145
institution Nanyang Technological University
language English
last_indexed 2024-10-01T05:37:12Z
publishDate 2022
record_format dspace
spelling ntu-10356/1631452022-11-25T02:13:47Z Meta-based self-training and re-weighting for aspect-based sentiment analysis He, Kai Mao, Rui Gong, Tieliang Li, Chen Cambria, Erik School of Computer Science and Engineering Engineering::Computer science and engineering Aspect-Based Sentiment Analysis Meta Learning Aspect-based sentiment analysis (ABSA) means to identify fine-grained aspects, opinions, and sentiment polarities. Recent ABSA research focuses on utilizing multi-task learning (MTL) to achieve less computational costs and better performance. However, there are certain limits in MTL-based ABSA. For example, unbalanced labels and sub-task learning difficulties may result in the biases that some labels and sub-tasks are overfitting, while the others are underfitting. To address these issues, inspired by neuro-symbolic learning systems, we propose a meta-based self-training method with a meta-weighter (MSM). We believe that a generalizable model can be achieved by appropriate symbolic representation selection (in-domain knowledge) and effective learning control (regulation) in a neural system. Thus, MSM trains a teacher model to generate in-domain knowledge (e.g., unlabeled data selection and pseudo-label generation), where the generated pseudo-labels are used by a student model for supervised learning. Then, the meta-weighter of MSM is jointly trained with the student model to provide each instance with sub-task-specific weights to coordinate their convergence rates, balancing class labels, and alleviating noise impacts introduced from self-training. The following experiments indicate that MSM can utilize 50% labeled data to achieve comparable results to state-of-arts models in ABSA and outperform them with all labeled data. This work has been supported by grant Key Research and Development Program of Ningxia Hui Nationality Autonomous Region (2022BEG02025); grant Key Research and Development Program of Shaanxi Province (2021GXLH-Z095); grant RIE2020 Industry Alignment Fund aˆ Industry Collaboration Projects (IAF-ICP) Funding Initiative; grant 61721002 from the Innovative Research Group of the National Natural Science Foundation of China, and grant IRT 17R86 from the Innovation Research Team of the Ministry of Education, Project of China Knowledge Centre for Engineering Science and Technology. 2022-11-25T02:13:47Z 2022-11-25T02:13:47Z 2022 Journal Article He, K., Mao, R., Gong, T., Li, C. & Cambria, E. (2022). Meta-based self-training and re-weighting for aspect-based sentiment analysis. IEEE Transactions On Affective Computing, 3202831-. https://dx.doi.org/10.1109/TAFFC.2022.3202831 1949-3045 https://hdl.handle.net/10356/163145 10.1109/TAFFC.2022.3202831 2-s2.0-85137543873 3202831 en IEEE Transactions on Affective Computing © 2022 IEEE. All rights reserved.
spellingShingle Engineering::Computer science and engineering
Aspect-Based Sentiment Analysis
Meta Learning
He, Kai
Mao, Rui
Gong, Tieliang
Li, Chen
Cambria, Erik
Meta-based self-training and re-weighting for aspect-based sentiment analysis
title Meta-based self-training and re-weighting for aspect-based sentiment analysis
title_full Meta-based self-training and re-weighting for aspect-based sentiment analysis
title_fullStr Meta-based self-training and re-weighting for aspect-based sentiment analysis
title_full_unstemmed Meta-based self-training and re-weighting for aspect-based sentiment analysis
title_short Meta-based self-training and re-weighting for aspect-based sentiment analysis
title_sort meta based self training and re weighting for aspect based sentiment analysis
topic Engineering::Computer science and engineering
Aspect-Based Sentiment Analysis
Meta Learning
url https://hdl.handle.net/10356/163145
work_keys_str_mv AT hekai metabasedselftrainingandreweightingforaspectbasedsentimentanalysis
AT maorui metabasedselftrainingandreweightingforaspectbasedsentimentanalysis
AT gongtieliang metabasedselftrainingandreweightingforaspectbasedsentimentanalysis
AT lichen metabasedselftrainingandreweightingforaspectbasedsentimentanalysis
AT cambriaerik metabasedselftrainingandreweightingforaspectbasedsentimentanalysis