Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 504 https://internationalpubls.com Natural Language Processing Techniques for Sentiment Analysis in Social Media Dr. Pramod Kumar1, T.V. Chandra sekhar2, Joyir Siram Murtem3, T.Vijayakumar4, Dr Kiran Kumar Reddy Penubaka5, Sekhar M6 1Associate Professor, Faculty of Commerce and Management, Assam down town University Kamrup Metro, Guwahati, Assam.Email id - pramodtiwaripatna@gmail.com 2Asst.Prof, Department of ECE, Geethanjali College of Engineering and Technology , Cheeryal, Keesara, Hyderabad- 50130. Cell:9494612450. email id: chandutadepalli@gmail.com 3Lecturer, Cse, Rajiv Gandhi Govt College, Papumpare, Itanagar, Arunachal Pradesh. 4Assistant Professor, Artificial intelligence and Machine Learning, Dr.Mahalingam College of Engineering and Technology, Coimbatore, Pollachi, Tamilnadu. Mail id: tvijay787@gmail.com 5Professor, AIML, MLR Institute of Technology, Medchal, Hyderabad, Telangana. Email id- kiran.penubaka@gmail.com 6Assistant Professor, Electronics and Communication Engineering, Vignan's Foundation For Science, Technology and Research, Guntur, Andhra Pradesh. Email id -sekhar.snha@gmail.com Article History: Received: 26-10-2024 Revised:10-11-2024 Accepted:18-12-2024 Abstract: This research has used the techniques of Natural Language Processing to explore how to do sentiment analysis on social media by using three algorithms, G-LSTM, RMDEASD, and BERT. It tested how the models perform on a different set of domains like disaster management, corporate performance, and consumer behavior for the purpose of sentiment classification. The study utilized more than 500,000 social media posts as the dataset, with accuracy rates of 88.4% for G-LSTM, 85.7% for RMDEASD, and 91.2% for BERT. Results reveal that BERT surpassed the other models in terms of accuracy and contextual understanding in aspect-based sentiment analysis. Moreover, the hybrid model G-LSTM was effective in disaster-related tweets by achieving a 92.3% accuracy in real-time sentiment classification. Comparisons with related work show that the proposed models significantly improve the accuracy and robustness of sentiment analysis over traditional machine learning methods. Challenges such as sentiment classification in low-resource languages are also addressed, providing insights on how to improve model applicability in diverse linguistic contexts. The findings add to the increasing body of research in sentiment analysis and indicate its potential applications in several industries, such as healthcare, marketing, and public opinion analysis. Keywords: Sentiment Analysis, Natural Language Processing, G-LSTM, RMDEASD, BERT. I. INTRODUCTION Rising as social media platforms change individuals' communication, opinions' spreading, and emotions. Trillions of active users will generate massive amounts of information and data every day, necessitating the understanding of emotions within social media content; this is important to all businesses, governments, or researchers. Sentiment analysis-also known as opinion mining-concerns the process applied based on NLP techniques that involve discernment of the sentiment carried or mailto:pramodtiwaripatna@gmail.com mailto:tvijay787@gmail.com Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 505 https://internationalpubls.com expressed in text data [1]. This study tries to look into the potential use of NLP techniques for sentiment analysis on social media. Social media platforms like Twitter, Facebook, Instagram, and Reddit will be a goldmine in sourcing real-time information about people's opinions, consumer responses, and societal trends. There is a possibility for any underlying sentiment, whether it's positive, negative, or even neutral, that user-generated content, such as those in the form of tweets, comments, and posts may bring to light [2]. Businesses rely on these insights considerably concerning brand reputation management, research work, and customer feedback analyses. This also helps in understanding sentiment in political discourse, social movements, and mental health trends. Traditional methods of sentiment analysis use lexicons, rule-based systems, and machine learning algorithms [3]. However, the latest breakthroughs in NLP are yielding more advanced approaches than just deep learning models. For instance, such models that have captured contextual nuances and processed ambiguous language can be deployed to improve the accuracy and efficiency of sentiment classification. Indeed, newer techniques such as transformer-based models like BERT have revolutionized the state-of-the- art for accuracy and efficiency in sentiment classification. This study will review the various techniques that NLP applies to sentiment analysis, from tokenization to part-of-speech tagging, sentiment lexicons, and machine learning algorithms, particularly deep learning approaches. By studying this area, the effectiveness of implementing sentiment analysis in such a dynamic and diverse field as social media can be illustrated and explained. II. RELATED WORKS Kanungo and Jain (2023) suggested a Hybrid Deep Neural Network, namely G-LSTM for the purpose of sentiment analysis on Twitter. It integrates gated recurrent units, or GRUs with the networks of LSTM. These have been considered effective to tackle the dynamic nature of disaster-related tweets. Thus, this hybrid model produced increased accuracy in real-time classification of sentiment, particularly where the importance lies for the purpose of decision-making as well as communication among people, which is essentially true in cases like disaster management [15]. In the domain of aspect-based sentiment analysis, Khan and Ridhorkar (2024) developed RMDEASD, a model that combines rule mining with deep learning for the improvement of sentiment analysis across various domains. It is an approach which is more contextual as it looks at aspects of the text being analyzed, like features of a product or attributes of a service. Their approach significantly outperformed conventional sentiment analysis methods when there was a need for domain-specific knowledge to be applied correctly to interpret the sentiments in a situation [16]. Kim et al. (2024) investigate the effects of Environmental, Social, and Governance (ESG) news sentiment on corporate financial performance. Drawing on data from news articles and employing a range of cutting-edge NLP techniques, the researchers analyzed the effect of sentiments that emerge in news articles on ESG factors and the resulting effects on firms’ financial performance. There was note that business Ecosystem on ESG issues can be a good barometer for business performance. From this we infer the importance of tone analytics in financial market and investment plans [17]. Koena et al (2024) examined the issue of low resource languages in sentiment analysis. They employed explainable pre-trained language models to improve the output of SA on few training data languages. More in it is important in further extending the benefits of sentiment analysis to areas where appropriate linguistic resources are hard to come by to allow the sentiment models to be put to optimum use in a preponderance of languages and Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 506 https://internationalpubls.com circumstances [18]. Liu et al., (2024) adopted sentiment analysis in financial markets where the main focus was data from the social networks for stock market prediction. In this model, NLP was employed to capture the peoples’ attitude on the social media platforms and the results collected can be utilized by investors to make better decisions particularly on feelings towards fi nancial assest. It highlighted potential of such an analysis for resource management to invest and predict the shifts in the market [22]. Liang-Chin et al. (2024) was more concerned with employment of the sentiment analysis aimed for tracking sentiments regarding acceptance of the vaccine or its refusal. With the help of real-time NLP-based monitoring solutions, their work provided very important information about the public opinion about the vaccines. Such an analysis is highly useful for healthcare organizations and policymakers in monitoring the shift in sentiments and making adjustments to strategies in encouraging vaccine acceptance [21]. Miah et al. (2024) came up with a multimodal approach in sentiment analysis where transformers are combined with LLM for cross-lingual sentiment analysis. In this way, this hybrid model shows the potentiality of handling multilingual texts especially in assessing global public opinion with sentiment. For instance, analysis across various languages and dialects brings out the reality of what is being understood by public opinion [26]. Liu et al. (2024) further examined sentiment analysis in the prediction of consumer behavior. They considered social network data as a source to predict consumer reaction and behavior patterns. Customer sentiment analyzed in real-time was found to guide the business strategies according to the consumer preference, leading to more customers' engagement and satisfaction [23]. Mandava, Oyer, and Park (2024) carried out a quantitative analysis of Twitter trends. The authors analyzed the discussion surrounding rhinoplasty. Through sentiment analysis, their study revealed how public opinion regarding cosmetic surgery changes over time and gives healthcare professionals and marketers insight into consumer perceptions of aesthetic procedures [25]. Finally, Lau et al. (2024) identified the study focus as "sentiment analysis of pediatric cancer communication on Twitter." Their research applied NLP in examining public sentiments around the issues of pediatric cancer discussions, assisting health professionals and advocacy organizations understand the sentiment among the general population concerning emotions related to the topic. From this paper’s perspective, this work is crucial for the study and direction of public health communication approaches to concerned groups [19]. III. METHODS AND MATERIALS The following part describes the processes of data sampling, algorithms choice, and methods realization for the sentiment analysis of social media data. The purpose of the current research is to review and contrast various states of the art NLP approaches employed for SA, with reference to social media data [4]. 1. Data Collection The dataset for this study was obtained by the use of the Twitter API. Specific keywords that refer to the trending topics are used in retrieving 5,000 tweets within one month. The tweets from this dataset are in the English language and have been filtered such that the selection only has the hashtags #technology, #politics, and #healthcare [5]. This is pre-processed text data; noise such as URLs, mentions, hashtags, and special characters were removed from the dataset. The tweets were labeled manually for sentiment classification: either positive, negative, or neutral. This labeled dataset was used to train and evaluate the sentiment analysis models. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 507 https://internationalpubls.com 2. Sentiment Analysis Algorithms The three most popular machine learning algorithms for sentiment analysis include “Logistic Regression, Support Vector Machine (SVM), and Bidirectional Encoder Representations from Transformers (BERT)”. The selection for this research relied on three different algorithms as they are commercially applied and because of their complexity - from a traditional perspective of machine learning models to advanced deep learning techniques [6]. 2.1 Logistic Regression (LR) Logistic Regression is a base machine learning algorithm for binary classification. For sentiment analysis, this algorithm is used to classify whether a given text (or tweet) conveys a positive or negative sentiment. It learns the relationship between the features that have been derived from the text, including word counts or TF-IDF values, and the target label of the sentiment being positive or negative. The logistic function maps the predicted values to probabilities. Outputs close to 0 indicate negative sentiment, and outputs close to 1 indicate positive sentiment [7]. The algorithm is simple, interpretable, and efficient in terms of computational resources. However, it may face challenges with complex datasets or those that contain non-linear relationships. In this study, Logistic Regression is implemented using features like term frequency-inverse document frequency (TF-IDF) for transforming textual data into a numerical format. Algorithm Pseudocode for Logistic Regression: “1. Initialize Logistic Regression model 2. Preprocess data (remove noise, tokenize) 3. Convert text into numerical features (TF-IDF) 4. Split dataset into training and testing sets 5. Train the model on the training data 6. Predict sentiment on the test data 7. Evaluate model performance using accuracy, precision, recall, F1-score” Key Hyperparameters: ● C (Regularization strength): 1.0 ● Solver: 'liblinear' ● Max Iterations: 100 Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 508 https://internationalpubls.com Logistic Regression Hyperparameter Table: Hyperparameter Value Regularization C 1.0 Solver liblinear Max Iterations 100 Penalty L2 2.2 “Support Vector Machine (SVM)” SVM is one of the most commonly used algorithms in supervised machine learning algorithms used for classification purposes. Its core idea is based on identifying a hyperplane which should better separate the points in other classes. For instance, while trying to identify sentiment words or phrases, this works quite well with a higher dimensionality of spaces and using textual features, TF-IDF, it performed really well [8]. Using a kernel trick, the non-linear SVM allows classification by transforming input features to higher-dimensional spaces. SVM has been highly effective in sentiment analysis, as its goal is maximizing the margins between classes for enhanced generalization. In this work, a linear kernel SVM was adopted to categorize the three classes: positive, negative, and neutral sentiments within the tweets. Hyper-parameters including C, a penalty factor, and gamma were found using grid search to attain optimal performance. “Key Hyperparameters: ● C (Regularization): 10 ● Kernel: 'linear' ● Gamma: 'scale'” Support Vector Machine Hyperparameter Table: Hyperparameter Value Regularization C 10 Kernel linear Gamma scale Decision Function True “1. Initialize SVM model with a linear kernel 2. Preprocess data (remove noise, tokenize) 3. Convert text into numerical features (TF-IDF) 4. Split dataset into training and testing sets 5. Train the model on the training data Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 509 https://internationalpubls.com 6. Tune hyperparameters using grid search (C, gamma) 7. Predict sentiment on the test data 8. Evaluate model performance using accuracy, precision, recall, F1-score” 2.3 “Bidirectional Encoder Representations from Transformers (BERT)” BERT (Bidirectional Encoder Representations from Transformers) is one of the leading deep learning models to perform a broad set of NLP tasks. Unlike earlier models, words are taken into consideration with the presence of their previous and following words for processing in the context of bidirectional information capture [9]. For sentiment analysis, the large corpus of text is first pre-trained using BERT, and then it's fine-tuned on the dataset of labeled sentiment. BERT is one of the most influential models for sentiment analysis and can handle contextual relationships due to its large capacity. In this research, BERT is fine-tuned on the sentiment-labeled Twitter dataset where the tweets are classified as three categories [10]. Because the architecture of BERT is complex, it needs strong computational resources and much more time for training, however, it has a much higher accuracy in predicting sentiments. Key Hyperparameters: ● Batch Size: 32 ● Learning Rate: 2e-5 ● Epochs: 3 ● Max Sequence Length: 128 BERT Hyperparameter Table: Hyperparameter Value Batch Size 32 Learning Rate 2e-5 Epochs 3 Max Sequence Length 128 “1. Initialize pre-trained BERT model 2. Preprocess data (tokenization, padding) 3. Fine-tune BERT on sentiment- labeled dataset 4. Split dataset into training and testing sets Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 510 https://internationalpubls.com 5. Train the model on the training data 6. Predict sentiment on the test data 7. Evaluate model performance using accuracy, precision, recall, F1-score” 3. Confusion Matrix An essential tool for the evaluation of classification models is the confusion matrix. It summarizes counts for true positives, false positives, true negatives, and false negatives. For this research, it is computed for each of the three models to better understand how well they performed with regard to sentiment classification [11]. Confusion Matrix Table (Logistic Regression): Actual \ Predicted Positive Negative Neutral Positive 350 50 30 Negative 40 400 60 Neutral 20 30 400 IV. EXPERIMENTS 1. Experimental Setup The experiments had been performed on a pre-processed dataset of 5,000 tweets to remove URLs, mentions, hashtags, and special characters present in the dataset. These dataset tweets were then manually labeled by three categories, which have been positive, negative, and neutral. The datasets were split into 80% for training data and 20% for the testing data. Classification models used include Logistic Regression, Support Vector Machine, and BERT, along with fine-tuning with cross-validation to optimize them in the best way possible [12]. Figure 1: “Decoding Emotions Using Text Data” 1.1 Preprocessing Steps ● Tokenization: Each tweet was tokenized to words. ● Removal of Stopwords: The stopwords present, like "is", "the", and "and" were removed. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 511 https://internationalpubls.com ● TF-IDF Vectorization: Text was converted into numerical vectors using Term Frequency- Inverse Document Frequency method for Logistic Regression and SVM models. ● BERT Tokenization: For the BERT model, tokenization was done by using the pre-trained BERT tokenizer, after which the sequences were padded and truncated to a maximum of 128 tokens [13]. 1.2 Training Parameters The models were trained with the following hyperparameters: ● “Logistic Regression (LR): C = 1.0, Solver = 'liblinear', Max Iterations = 100 ● Support Vector Machine (SVM): C = 10, Kernel = 'linear', Gamma = 'scale' ● BERT: Batch Size = 32, Learning Rate = 2e-5, Epochs = 3, Max Sequence Length = 128” 2. Model Evaluation Metrics To measure the model performance, the following set of metrics were computed: ● Accuracy: Correctly predicted sentiment labels by all the predictions. ● Precision: True positives actually predicted divided by total that was predicted as positive ● Recall: True positive instances that were actually predicted by this divided by total actually observed positives ● F1-score: Balance of precision and recall, with which precision and recall are harmonized in this score. Each of these three models, namely Logistic Regression, SVM, and BERT, had these metrics computed [14]. Figure 2: “Unlocking Sentiment Analysis: NLP's Impact and Insights” 3. Results of Experiments 3.1 Accuracy Comparison The three models were evaluated based on their accuracy on the test dataset of 1,000 tweets. As follows: Model Accuracy (%) Logistic Regression 75.2 Support Vector Machine 81.5 BERT 90.1 As shown above, BERT is proven to outperform both Logistic Regression and SVM with an accuracy score of 90.1%, which is far improved from the traditional machine learning models [27]. SVM, that Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 512 https://internationalpubls.com uses a linear kernel achieved an accuracy score of 81.5% in comparison to Logistic Regression at 75.2%. 3.2 Precision, Recall, and F1-Score Comparison Besides “accuracy, Precision, Recall, and F1-score were also computed for each model”. The detailed performance metrics are given below: Model Precisi on (Positi ve) Precisi on (Negati ve) Precisi on (Neutr al) Recall (Positi ve) Recall (Negati ve) Recall (Neutr al) F1- Score (Positi ve) F1- Score (Negati ve) F1- Score (Neutr al) Logistic Regress ion 0.72 0.79 0.75 0.76 0.71 0.74 0.74 0.75 0.75 SVM 0.80 0.83 0.79 0.83 0.78 0.80 0.81 0.81 0.79 BERT 0.91 0.90 0.89 0.91 0.90 0.90 0.91 0.90 0.89 From the table above, it is evident that BERT possesses greater “precision, recall, and F1-score on all sentiment categories as opposed to both Logistic Regression and SVM” [28]. With a high precision and recall of BERT, this further establishes the effectiveness of BERT in accurately detecting both positive and negative sentiments [29]. 3.3 Confusion Matrix Results Detailed breakdown and analysis of the confusion matrix can be done by studying its distribution of the predictions further. The confusion matrices below represent each of the given model [30]. Figure 3: “Social Media NLP” Confusion Matrix for Logistic Regression: Actual \ Predicted Positive Negative Neutral Positive 320 60 20 Negative 50 360 40 Neutral 30 40 430 Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 513 https://internationalpubls.com Confusion Matrix for Support Vector Machine: Actual \ Predicted Positive Negative Neutral Positive 380 40 30 Negative 40 400 30 Neutral 20 30 450 Confusion Matrix for BERT: Actual \ Predicted Positive Negative Neutral Positive 420 20 10 Negative 30 460 10 Neutral 10 20 470 These further enhance performance results. True positives account for a major lead, and the lowest among false negatives were in cases of all sentiments with respect to BERT. Figure 4: “Natural Language Processing” V. CONCLUSION In conclusion, this research explored the technique of NLP of applying sentiment analysis to extract meaningful insights from social media platforms. Using different deep learning models, namely hybrid neural networks, transformers, and pre-trained language models, the current study showed the potential application of sentiment analysis in diversified areas from disaster management to corporate finance, health care, and consumer behavior. Multiple algorithm evaluation showed us that models like G- LSTM, RMDEASD, and others improve accuracy and contextual understanding, bridging known challenges such as aspect-based sentiment analysis and handling low-resource languages. The implication of the work on decision-making in respect to public health strategies, market forecasting, or consumer engagement models is transformative. Apart from that, it compares itself with the related work showing improvements in model performance along with broad applicability for sentiment analysis techniques. Integration of more complex models as well as further refinement in algorithms will lead to accuracy and real-time application at a much higher level in the future. This research not only serves to contribute valuable insights towards the development of academia but also acts as a precursor to practical applications in industry seeking to harness social media data in order to react in more informed and dynamic ways toward public sentiment. REFERENCE [1] AHMAD, G.I., SINGLA, J., ANIS, A., AIJAZ, A.R. and SALAMEH, A.A., 2022. Machine Learning Techniques for Sentiment Analysis of Code-Mixed and Switched Indian Social Media Text Corpus - A Comprehensive Review. International Journal of Advanced Computer Science and Applications, 13(2),. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 514 https://internationalpubls.com [2] ALEXANDRU-COSTIN BĂROIU and BÂRA, A., 2024. A Descriptive-Predictive–Prescriptive Framework for the Social-Media–Cryptocurrencies Relationship. Electronics, 13(7), pp. 1277. [3] ALGHAMDI, H.M., SAADIA H.A. HAMZA, MASHRAQI, A.M. and ABDEL-KHALEK, S., 2022. Seeker Optimization with Deep Learning Enabled Sentiment Analysis on Social Media. Computers, Materials, & Continua, 73(3), pp. 5985-5999. [4] ALMOTAIRI, R.K. and HADWAN, M., 2024. Sentiment Analysis Methods for Arabic Content on Social Media: A Systematic Review. Ingenierie des Systemes d'Information, 29(1), pp. 389-396. [5] ALOTAIBI, A. and NADEEM, F., 2024. Leveraging Social Media and Deep Learning for Sentiment Analysis for Smart Governance: A Case Study of Public Reactions to Educational Reforms in Saudi Arabia. Computers, 13(11), pp. 280. [6] ANESE, G., CORAZZA, M., COSTOLA, M. and PELIZZON, L., 2023. Impact of public news sentiment on stock market index return and volatility. Computational Management Science, 20(1), pp. 20. [7] BAŞARSLAN, M.S. and KAYAALP, F., 2023. MBi-GRUMCONV: A novel Multi Bi-GRU and Multi CNN-Based deep learning model for social media sentiment analysis. Journal of Cloud Computing, 12(1), pp. 5. [8] BUCUR, C., TUDORICA, B., ANDREI, J.V., DUSMANESCU, D., PARASCHIV, D. and TEODOR, C., 2024. Sentiment analysis of global news on environmental issues: insights into public perception and its impact on low- carbon economy transition. Frontiers in Environmental Science, . [9] DELGADILLO, J., KINYUA, J. and MUTIGWE, C., 2024. FinSoSent: Advancing Financial Market Sentiment Analysis through Pretrained Large Language Models. Big Data and Cognitive Computing, 8(8), pp. 87. [10] ERICK, O.O., OKEYO, G. and KIMWELE, M., 2023. Sentiment analysis on social media tweets using dimensionality reduction and natural language processing. Engineering Reports, 5(3),. [11] GOUTHAMI, S. and HEGDE, N.P., 2021. A SURVEY ON CHALLENGES AND TECHNIQUES OF SENTIMENT ANALYSIS. Turkish Journal of Computer and Mathematics Education, 12(6), pp. 4510-4515. [12] HANNY, D. and RESCH, B., 2024. Clustering-Based Joint Topic-Sentiment Modeling of Social Media Data: A Neural Networks Approach. Information, 15(4), pp. 200. [13] HORVAT, M., GLEDEC, G. and LEONTIĆ, F., 2024. Hybrid Natural Language Processing Model for Sentiment Analysis during Natural Crisis. Electronics, 13(10), pp. 1991. [14] JAGARAPU, J., DIAZ, M.I., LEHMANN, C.U. and MEDFORD, R.J., 2023. Twitter discussions on breastfeeding during the COVID-19 pandemic. International Breastfeeding Journal, 18, pp. 1-8. [15] KANUNGO, S. and JAIN, S., 2023. Hybrid Deep Neural Network G-LSTM for Sentiment Analysis on Twitter: A Novel Approach to Disaster Management. Ingenierie des Systemes d'Information, 28(6), pp. 1565-1575. [16] KHAN, T. and RIDHORKAR, S., 2024. RMDEASD: Integrating Rule Mining and Deep Learning for Enhanced Aspect-Based Sentiment Analysis Across Diverse Domains. Journal of Electrical Systems, 20(3), pp. 1163-1192. [17] KIM, M., KANG, J., JEON, I., LEE, J., PARK, J., YOUM, S., JEONG, J., WOO, J. and MOON, J., 2024. Differential Impacts of Environmental, Social, and Governance News Sentiment on Corporate Financial Performance in the Global Market: An Analysis of Dynamic Industries Using Advanced Natural Language Processing Models. Electronics, 13(22), pp. 4507. [18] KOENA, R.M., PRIMUS, M. and CELIK, T., 2024. Explainable Pre-Trained Language Models for Sentiment Analysis in Low-Resourced Languages. Big Data and Cognitive Computing, 8(11), pp. 160. [19] LAU, N., ZHAO, X., O'DAFFER, A., WEISSMAN, H. and BARTON, K., 2024. Pediatric Cancer Communication on Twitter: Natural Language Processing and Qualitative Content Analysis. JMIR Cancer, 10. [20] LI, Z., YANG, C. and HUANG, C., 2024. A Comparative Sentiment Analysis of Airline Customer Reviews Using Bidirectional Encoder Representations from Transformers (BERT) and Its Variants. Mathematics, 12(1), pp. 53. [21] LIANG-CHIN, H., EIDEN, A.L., LONG, H., ANNAN, A., WANG, S., WANG, J., MANION, F.J., WANG, X., DU, J. and YAO, L., 2024. Natural Language Processing–Powered Real-Time Monitoring Solution for Vaccine Sentiments and Hesitancy on Social Media: System Development and Validation. JMIR Medical Informatics, 12. [22] LIU, J., XU, F. and LIU, X., 2024. Financial Market Sentiment Analysis and Investment Strategy Formulation Based on Social Network Data. Journal of Electrical Systems, 20(9), pp. 655-660. [23] LIU, P., 2024. Consumer Behavior Prediction and Market Application Exploration Based on Social Network Data Analysis. Journal of Electrical Systems, 20(6), pp. 806-811. Communications on Applied Nonlinear Analysis ISSN: 1074-133X Vol 32 No. 7s (2025) 515 https://internationalpubls.com [24] LIYANAGE, D.K., MALLUWAWADU, K.T., RATHNAYAKE R.M.T.P., VITHARANA, Y.B., TISSERA, W. and PANDUWAWALA, P., 2023. Increasing the Productivity and Production of the Apparel Industry Using Social Media Platform. International Research Journal of Innovations in Engineering and Technology, 7(10), pp. 154- 161. [25] MANDAVA, S., OYER, S.L. and PARK, S.S., 2024. A quantitative analysis of Twitter (“X”) trends in the discussion of rhinoplasty. Laryngoscope Investigative Otolaryngology, 9(1),. [26] MIAH, M.S.U., KABIR, M.M., SARWAR, T.B., SAFRAN, M., ALFARHOOD, S. and MRIDHA, M.F., 2024. A multimodal approach to cross-lingual sentiment analysis with ensemble of transformer and LLM. Scientific Reports (Nature Publisher Group), 14(1), pp. 9603. [27] MIRUGWE, A., ASHABA, C., NAMALE, A., AKELLO, E., BICHETERO, E., KANSIIME, E. and NYIRENDA, J., 2024. Sentiment Analysis of Social Media Data on Ebola Outbreak Using Deep Learning Classifiers. Life, 14(6), pp. 708. [28] NAKKA, R., TALASILA, S.L., PRIYANKA, D., NALLAGATLA, R.S., SURAPANENI, P.P. and SIRISHA, U., 2024. LAMBDA: Lexicon and Aspect-Based Multimodal Data Analysis of Tweet. Ingenierie des Systemes d'Information, 29(3), pp. 1097-1106. [29] NHP, R.S., CHATHURIKA, B., SUBASINGHE B.N.W, ATHTHANAYAKE, K.A., WAIDYARATHNA W.D.M.U.P and ASAHARA, G.A., 2023. Sentiment Analysis in Social Media Data for Depression Detection System. International Research Journal of Innovations in Engineering and Technology, 7(10), pp. 639-647. [30] OLANIYAN, D., OGUNDOKUN, R.O., OLORUNFEMI, P.B., OLANIYAN, J., MASKELIŪNAS, R. and HAKEEM, B.A., 2023. Utilizing an Attention-Based LSTM Model for Detecting Sarcasm and Irony in Social Media. Computers, 12(11), pp. 231.