Cover Image

Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach

In aspect-based sentiment analysis, tasks are diverse and consist of aspect term extraction, aspect categorization, opinion term extraction, sentiment polarity classification, and relation extractions of aspect and opinion terms. These tasks are generally carried out sequentially using more than one...

Full description

Saved in:
Bibliographic Details
Main Authors: Ahmad Genadi, Rifo, Khodra, Masayu Leylia
Format: Article
Language:English
Published: Universiti Utara Malaysia Press 2022
Subjects:
TK Electrical engineering. Electronics Nuclear engineering
Online Access:https://repo.uum.edu.my/id/eprint/28805/1/JICT%2021%2002%202022%20255-277.pdf
https://repo.uum.edu.my/id/eprint/28805/
Tags: Add Tag
No Tags, Be the first to tag this record!
id my.uum.repo.28805
record_format eprints
spelling my.uum.repo.288052023-02-01T00:55:16Z https://repo.uum.edu.my/id/eprint/28805/ Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach Ahmad Genadi, Rifo Khodra, Masayu Leylia TK Electrical engineering. Electronics Nuclear engineering In aspect-based sentiment analysis, tasks are diverse and consist of aspect term extraction, aspect categorization, opinion term extraction, sentiment polarity classification, and relation extractions of aspect and opinion terms. These tasks are generally carried out sequentially using more than one model. However, this approach is inefficient and likely to reduce the model’s performance due to cumulative errors in previous processes. The co-extraction approach with Dual crOss-sharEd RNN (DOER) and span-based multitask acquired better performance than the pipelined approaches in English review data. Therefore, this research focuses on adapting the co-extraction approach where the extraction of aspect terms, opinion terms, and sentiment polarity are conducted simultaneously from review texts. The co-extraction approach was adapted by modifying the original frameworks to perform unhandled subtask to get the opinion triplet. Furthermore, the output layer on these frameworks was modified and trained using a collection of Indonesian-language hotel reviews. The adaptation was conducted by testing the output layer topology for aspect and opinion term extraction as well as variations in the type of recurrent neural network cells and model hyperparameters used, and then analysing the results to obtain a conclusion. The two proposed frameworks were able to carry out opinion triplet extraction and achieve decent performance. The DOER framework achieves better performance than the baselines on aspect and opinion term extraction tasks. Universiti Utara Malaysia Press 2022 Article PeerReviewed application/pdf en cc4_by https://repo.uum.edu.my/id/eprint/28805/1/JICT%2021%2002%202022%20255-277.pdf Ahmad Genadi, Rifo and Khodra, Masayu Leylia (2022) Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach. Journal of Information and Communication Technology, 21 (02). pp. 255-277. ISSN 2180-3862
institution Universiti Utara Malaysia
building UUM Library
collection Institutional Repository
continent Asia
country Malaysia
content_provider Universiti Utara Malaysia
content_source UUM Institutional Repository
url_provider http://repo.uum.edu.my/
language English
topic TK Electrical engineering. Electronics Nuclear engineering
spellingShingle TK Electrical engineering. Electronics Nuclear engineering
Ahmad Genadi, Rifo
Khodra, Masayu Leylia
Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
description In aspect-based sentiment analysis, tasks are diverse and consist of aspect term extraction, aspect categorization, opinion term extraction, sentiment polarity classification, and relation extractions of aspect and opinion terms. These tasks are generally carried out sequentially using more than one model. However, this approach is inefficient and likely to reduce the model’s performance due to cumulative errors in previous processes. The co-extraction approach with Dual crOss-sharEd RNN (DOER) and span-based multitask acquired better performance than the pipelined approaches in English review data. Therefore, this research focuses on adapting the co-extraction approach where the extraction of aspect terms, opinion terms, and sentiment polarity are conducted simultaneously from review texts. The co-extraction approach was adapted by modifying the original frameworks to perform unhandled subtask to get the opinion triplet. Furthermore, the output layer on these frameworks was modified and trained using a collection of Indonesian-language hotel reviews. The adaptation was conducted by testing the output layer topology for aspect and opinion term extraction as well as variations in the type of recurrent neural network cells and model hyperparameters used, and then analysing the results to obtain a conclusion. The two proposed frameworks were able to carry out opinion triplet extraction and achieve decent performance. The DOER framework achieves better performance than the baselines on aspect and opinion term extraction tasks.
format Article
author Ahmad Genadi, Rifo
Khodra, Masayu Leylia
author_facet Ahmad Genadi, Rifo
Khodra, Masayu Leylia
author_sort Ahmad Genadi, Rifo
title Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
title_short Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
title_full Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
title_fullStr Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
title_full_unstemmed Opinion Triplet Extraction for Aspect-Based Sentiment Analysis Using Co-Extraction Approach
title_sort opinion triplet extraction for aspect-based sentiment analysis using co-extraction approach
publisher Universiti Utara Malaysia Press
publishDate 2022
url https://repo.uum.edu.my/id/eprint/28805/1/JICT%2021%2002%202022%20255-277.pdf
https://repo.uum.edu.my/id/eprint/28805/
_version_ 1756687043202646016
score 13.211869

Similar Items