Article
Computer Science, Artificial Intelligence
Jhe-Wei Lin, Tran Duy Thanh, Rong-Guey Chang
Summary: This paper proposes a method for improving sentiment analysis using multi-channel word embeddings, by improving word representation, applying attention mechanism, and using state-of-the-art deep learning models. The experimental results demonstrate that the proposed method achieves higher accuracy compared to baseline methods and highlights the importance of better word representation based on multi-channel pre-trained word embeddings.
Article
Computer Science, Artificial Intelligence
Leila Moudjari, Farah Benamara, Karima Akli-Astouati
Summary: This study proposes multi-level embeddings as a novel solution for in-depth investigation into the impact of various subwords configurations in Arabic social media text classification. By conducting extrinsic evaluations on sentiment, emotion, and irony detection tasks, the results show that our proposed multi-level embeddings outperform current static and contextualised embeddings as well as best performing state of the art models.
COMPUTER SPEECH AND LANGUAGE
(2021)
Article
Psychology, Multidisciplinary
Baitao Liu
Summary: This study focuses on the method of emotion analysis in the application of psychoanalysis based on sentiment recognition. The improved C-BiL model is applied to the sentiment recognition module, and it effectively realizes the function of sentiment recognition. The experimental results show that the C-BiL model designed in this study achieves relatively high accuracy in different datasets.
FRONTIERS IN PSYCHOLOGY
(2022)
Article
Computer Science, Artificial Intelligence
Hande Aka Uymaz, Senem Kumova Metin
Summary: Text is commonly used for natural language processing studies, but accurately reflecting meaning and detecting emotion pose challenges. Word embeddings, such as Word2Vec and GloVe, are frequently used to represent textual data and extract semantic information. However, these models may not capture emotive data effectively, leading to unexpected results in sentiment and emotion detection.
EXPERT SYSTEMS WITH APPLICATIONS
(2023)
Article
Chemistry, Multidisciplinary
Lifang Wu, Sinuo Deng, Heng Zhang, Ge Shi
Summary: Sentiment analysis is a challenging task due to the affective gap in accurately extracting sentimental features from visual contents. Previous approaches neglect the interaction among objects and may introduce noisy features. In this paper, a method called Sentiment Interaction Distillation (SID) Network is proposed to guide feature learning using object sentimental interaction. Experimental results show that the reasonable use of interaction features can improve the performance of sentiment analysis.
APPLIED SCIENCES-BASEL
(2022)
Article
Chemistry, Multidisciplinary
Lifang Wu, Heng Zhang, Sinuo Deng, Ge Shi, Xu Liu
Summary: By leveraging the interactive characteristics of objects in sentimental space based on a Graph Convolutional Network (GCN), a new sentiment analysis framework is proposed. Experimental results demonstrate that this method outperforms existing algorithms, showing that the rational use of interaction features can improve sentiment analysis performance.
APPLIED SCIENCES-BASEL
(2021)
Article
Computer Science, Artificial Intelligence
Ajwa Aslam, Allah Bux Sargano, Zulfiqar Habib
Summary: In recent years, there has been a growing interest in multimodal sentiment analysis and emotion recognition due to its practical applications. This paper introduces a novel framework called AMSAER to address the challenges of working with multiple modalities. The proposed framework achieves notable performance improvement in sentiment analysis and emotion classification compared to state-of-the-art methods, with accuracy reaching 85% and 93% respectively.
APPLIED SOFT COMPUTING
(2023)
Article
Computer Science, Theory & Methods
Wenchuan Mu, Kwan Hui Lim, Junhua Liu, Shanika Karunasekera, Lucia Falzon, Aaron Harwood
Summary: ClusTop is a clustering-based topic modeling algorithm that automatically determines the number and types of discussion topics without the need for tuning parameters. It captures syntactic meaning in tweets and outperforms traditional methods in terms of topic coherence and precision.
JOURNAL OF BIG DATA
(2022)
Review
Computer Science, Artificial Intelligence
Jun Liu, Shuang Zheng, Guangxia Xu, Mingwei Lin
Summary: The study extended the CBoW word vector model and proposed a cross-domain sentiment-aware word embedding learning model, which can capture both the sentiment information and domain relevance of a word. The experimental results demonstrate that the model has higher accuracy and Macro-F1 value when dealing with sentiment information.
INTERNATIONAL JOURNAL OF MACHINE LEARNING AND CYBERNETICS
(2021)
Article
Computer Science, Information Systems
Baiyu Yang, Donghong Han, Rui Zhou, Di Gao, Gang Wu
Summary: This paper introduces a model for sentiment classification task on specific aspects, achieving superior performance in capturing the correspondences between aspects and opinion words in a sentence.
INFORMATION SCIENCES
(2022)
Article
Computer Science, Artificial Intelligence
Serpil Aslan, Soner Kiziloluk, Eser Sert
Summary: COVID-19, a novel virus from the coronavirus family, has caused a global outbreak that resulted in over 5.5 million deaths. The pandemic has had significant psychological effects on people's mental states, leading to anxiety and fear. Social media platforms, particularly Twitter, have been used by millions of people worldwide to share their opinions on COVID-19. In this study, a Twitter sentiment analysis (TSA) approach called TSA-CNN-AOA was proposed and demonstrated to have a higher classification performance compared to other similar approaches.
NEURAL COMPUTING & APPLICATIONS
(2023)
Article
Chemistry, Multidisciplinary
James Mutinda, Waweru Mwangi, George Okeyo
Summary: Sentiment analysis is a crucial area of research in natural language processing, with various applications. However, accurate sentiment analysis requires robust text representation techniques. This paper introduces a sentiment classification model called LeBERT, which combines sentiment lexicon, N-grams, BERT, and CNN. Experimental results show that LeBERT outperforms existing models.
APPLIED SCIENCES-BASEL
(2023)
Review
Computer Science, Information Systems
Bhart Gupta, P. Prakasam, T. Velmurugan
Summary: Understanding the sentiment of articles and movies is becoming a major issue due to diverse opinions expressed in reviews. In this research, a novel binary sentiment classification method is proposed, incorporating BERT embeddings and deep learning models for accurate sentiment analysis. The proposed model outperforms existing models in binary sentiment classification analysis, achieving a testing accuracy of 93.89% and an AUC value of 0.9828 on the IMDB movie review dataset.
MULTIMEDIA TOOLS AND APPLICATIONS
(2022)
Article
Computer Science, Artificial Intelligence
Sajjad Shumaly, Mohsen Yazdinejad, Yanhui Guo
Summary: The study conducted sentiment analysis on a Persian website using fastText and CNN methods, achieving higher accuracy and independence from pre-processing compared to other research. By collecting reviews, creating word embeddings, and comparing multiple models, the research addressed the main issue in Persian sentiment analysis.
PEERJ COMPUTER SCIENCE
(2021)
Article
Physics, Multidisciplinary
Jinfeng Zhou, Xiaoqin Zeng, Yang Zou, Haoran Zhu
Summary: This study proposes a new CNN model based on residual network technology and attention mechanisms, which can extract more abundant multi-scale sentiment features and address the loss of locally detailed information, thereby enhancing the accuracy of sentiment classification.