Aspect Based Sentiment Analysis for Explicit and Implicit Aspects in Restaurant Review using Grammatical Rules, Hybrid Approach, and SentiCircle

Abdullah, Rachmad; Suhariyanto, Suhariyanto; Sarno, Riyanarto

doi:10.22266/ijies2021.1031.27

Cited by 9 publications

(11 citation statements)

References 23 publications

(28 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We carried out two tasks of ABSA: aspect category identification (ACI) and aspect sentiment classification (ASC). BiLSTM-BERT obtained the best performance for the ACI task with an accuracy of 60.06%, macro precision of indicates that the domain of documents used to train the word embedding model is also important for better representation of words [19]. Moreover, we found no indication that using a bidirectional can improve performance.…”

Section: Multi-task Learning Absamentioning

confidence: 64%

“…Several studies have implemented deep learning methods for ATE and ACI tasks [14][15][16]. In another way, several rulebased methods have been developed to deal with ATE task [17][18][19][20][21], where Part of Speech (POS) Tagging and dependency parsing was executed to identify terms related to an aspect. Recently, most studies on ABSA only focused on improving the performance of ASC tasks [14][15][16][17][18][19][20][21].…”

Section: Introductionmentioning

confidence: 99%

“…In another way, several rulebased methods have been developed to deal with ATE task [17][18][19][20][21], where Part of Speech (POS) Tagging and dependency parsing was executed to identify terms related to an aspect. Recently, most studies on ABSA only focused on improving the performance of ASC tasks [14][15][16][17][18][19][20][21]. Moreover, most previous ACI and ASC are two separate tasks and could not meet practical application needs [22].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Stance Detection on Tweets with Multi-task Aspect-based Sentiment: A Case Study of COVID-19 Vaccination

2022

IJIES

View full text Add to dashboard Cite

Public opinion analyses on Twitter conducted based on sentiment analysis cannot identify the author's stance regarding agreement or disagreement with a given target. Stance detection determines whether the author of a text is in favor, against, or neutral towards a target. However, stance detection based on text-only is less representative opinion, especially on a tweet, which is a short text with slightly contextual information. Therefore, more information is needed to represent the author's stance better. In previous research, most research on stance detection was carried out using simple sentiment information to measure the support to target. This study addresses multi-task aspect-based sentiment analysis (ABSA) and social features for stance detection based on deep learning models of BiGRU-BERT on tweets. Our contribution combines aspect-based sentiment information with features based on textual and contextual information that does not emerge directly from Twitter texts. ABSA approach can provide more accurate sentiment information at aspect level on tweets, which is possible contains multiple issues discussed. Aspect information on tweets can reflect the issue that influences the author's stance toward a target. Multi-task learning was applied to help improve the generalization performance of ABSA with simultaneous processes. We extracted social attributes and online behavioral features for contextual information. Since same community tends to have the same opinion towards a target, we applied a community detection task and combine with the Twitter social attributes. The proposed method has significantly improved evaluation metrics (>10%) than textual features only for stance detection on tweets.

show abstract

Section: Multi-task Learning Absamentioning

confidence: 64%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Stance Detection on Tweets with Multi-task Aspect-based Sentiment: A Case Study of COVID-19 Vaccination

2022

IJIES

View full text Add to dashboard Cite

show abstract

“…Comparing synonym candidates with parents, children, or root nodes in large shopping taxonomies increased the accuracy of classification included in the complex entities [12]. WordNet is used to get synonyms and antonyms for expanding the meaning of words [13]. Word list with term frequency-inverse cluster frequency (TF-ICF) was extended to overcome the out-of-vocabulary (OOV) problems [14].…”

Section: Synonym-based Feature Expansionmentioning

confidence: 99%

Synonym based feature expansion for Indonesian hate speech detection

Ghozali

Sungkono

Sarno

et al. 2023

IJECE

View full text Add to dashboard Cite

Online hate speech is one of the negative impacts of internet-based social media development. Hate speech occurs due to a lack of public understanding of criticism and hate speech. The Indonesian government has regulations regarding hate speech, and most of the existing research about hate speech only focuses on feature extraction and classification methods. Therefore, this paper proposes methods to identify hate speech before a crime occurs. This paper presents an approach to detect hate speech by expanding synonyms in word embedding and shows the classification comparison result between Word2Vec and FastText with bidirectional long short-term memory which are processed using synonym expanding process and without it. The goal is to classify hate speech and non-hate speech. The best accuracy result without the synonym expanding process is 0.90, and the expanding synonym process is 0.93.

show abstract

“…Word similarity process is a word extraction process to get the word similarity of aspect term and category. The most frequently used word similarity is semantic similarity [13,14,15]. Semantic similarity is token-dependent, which is generated to be used as input, namely terms.…”

Section: Introductionmentioning

confidence: 99%

Aspect-Based Sentiment Analysis for Hotel Review Using LDA, Semantic Similarity, and BERT

2022

IJIES

View full text Add to dashboard Cite

Hotel review is frequently used as a main input in sentiment analysis process. It aims at helping travellers easily find more accurate information about hotel aspects in selecting hotels for their journeys. Based on the review datasets, hotel service organizers may evaluate guests' responses towards the services provided by the hotels. Hotel organizers, in turn, may also know the hotel aspects which need improvement for the next experiences. The common problems are because the processed data do not focus on small scale so that wrong selection of terms from review document frequently appears. The problem that often arises is that the amount of data that is processed is not limited to a small scale, so there are often errors in taking terms from a review document. Meanwhile, these terms are the main input source used for the assessment of aspect categorization and aspect-based sentiment analysis. So, we need aspect categorization and aspect-based sentiment analysis methods that can work automatically on a large scale with good accuracy results. In this study, first, the results of the pre-processing were processed using TF-ICF to obtain terms from reviews based on aspect keyword variables in each hotel aspect category. Next, LDA was used to get the hidden topic of each term. The aim was to obtain better terms accuracy results. Then, the aspect categorization process was carried out using BERT embedding and semantic similarity with the aim of obtaining more significant differences in similarity results in each aspect category so that the determination of aspect categories from a review could be more accurate. The results of the aspect extraction obtained an evaluation of the aspect categorization for each precision 0.86, recall 0.92, and f1-measure 0.89. Furthermore, BERT sentiment analysis method is used in the aspect-based sentiment analysis process. Finally, the evaluation result of aspect-based sentiment analysis obtained for each precision, recall, and f-1 measure are 0.96, 0.98, and 0.97.

show abstract

Aspect Based Sentiment Analysis for Explicit and Implicit Aspects in Restaurant Review using Grammatical Rules, Hybrid Approach, and SentiCircle

Cited by 9 publications

References 23 publications

Stance Detection on Tweets with Multi-task Aspect-based Sentiment: A Case Study of COVID-19 Vaccination

Stance Detection on Tweets with Multi-task Aspect-based Sentiment: A Case Study of COVID-19 Vaccination

Synonym based feature expansion for Indonesian hate speech detection

Aspect-Based Sentiment Analysis for Hotel Review Using LDA, Semantic Similarity, and BERT

Contact Info

Product

Resources

About