2026· International journal of engineering and technology· 0 citations· 15 references
TL;DR
Comparing and analysing the performance of several machine learning algorithms on fine-grained sentiment classification problems to examine their suitability and shortcomings for use as models in sentiment analysis suggests large language models perform significantly worse on the 28-class classification task in zero-shot settings, suggesting that they are better suited for generative and open-ended emotional interaction than for standardized classification benchmarks.
Abstract
— By comparing and analysing the performance of several machine learning algorithms on fine-grained sentiment classification problems to examine their suitability and shortcomings for use as models in sentiment analysis. Sentiment analysis remains a prominent research area in Natural Language Processing (NLP). However, systematic comparisons of whether these methods demonstrate superiority in fine-grained classification settings have yet to be conducted. Based on the GoEmotions dataset that includes 28 sentiment classes as our experiment’s reference for evaluating four kinds of models: logistic regression, BiLSTM, BERT, and the large-scale language model Claude (claude-haiku-4-5). The experiments employ TF-IDF feature extraction, sequence encoding, pre-training with fine-tuning, and zero-shot and few-shot prompting strategies. Based on this experiment, the model’s generalisation capabilities improve as architectural depth increases. The accuracy of logistic regression was only 55.11%; BERT obtained the highest F1 score and performed the best overall, and bidirectional pre-trained representations were most valuable among other methods. However, large language models perform significantly worse on the 28-class classification task in zero-shot settings, suggesting that they are better suited for generative and open-ended emotional interaction than for standardized classification benchmarks.
The paper presents a comparative analysis of the effectiveness of various text vectorization methods for the task of Sentiment Analysis of Russian-language reviews. The study covers classical frequency-based approaches (TF IDF, n-grams), statistical models (Word2Vec, FastText), and a contextual method based on the pre-trained BERT language model. The practical part of the research includes the implementation of text processing and classification pipelines using logistic regression and Naive Bayes classifiers. Experiments are conducted on a dataset of Russian-language reviews from a marketplace. Key comparison metrics are classification accuracy (accuracy, F1-score) and model training/inference time. The results show that on small datasets, classical methods with linear models demonstrate competitive quality with significantly lower computational costs. Contextual BERT embeddings show the best quality on the test set; however, their use is justified only with sufficient data volumes and the absence of strict real-time inference constraints. Based on the analysis, recommendations are given for choosing a vectorization method depending on the data volume and system performance requirements in real time.
O. I. Zakharova, S. Bednyak, Yaroslav Dmitrievich Kanunnikov· Infokommunikacionnye tehnolo...· 0 citations
Sentiment analysis has become an important task in natural language processing for understanding public opinions expressed in online reviews. However, most publicly available IMDb datasets are limited to binary sentiment labels, which restricts the ability of sentiment analysis systems to capture neutral opinions. This study proposes an efficient sentiment analysis framework that transforms the binary IMDb dataset into a three-class sentiment classification problem consisting of positive, neutral, and negative sentiments. The proposed approach integrates pseudolabeling with Parameter-Efficient Fine-Tuning (PEFT) using the Low-Rank Adaptation (LoRA) technique on the Longformer architecture. Experimental results show that the model achieves an accuracy of 77.06%, a weighted F1-score of 72.17%, and a Matthews Correlation Coefficient (MCC) of 0.6232. The results demonstrate that LoRA-based fine-tuning can significantly reduce computational requirements while maintaining competitive performance in sentiment classification tasks. These findings indicate that the proposed framework provides a practical and computationally efficient solution for large-scale sentiment analysis, particularly for environments with limited computational resources.
P. Hiskiawan, Wendy Tjung, Dustin Darmawan Isya Widjaja et al.· JRST: Jurnal Riset Sains dan...· 0 citations
Sentiment analysis is widely used to extract opinions from textual data; however, its application to morphologically rich languages such as Turkish remains challenging. This study investigates the optimization of classical machine learning classifiers and ensemble learning strategies for binary Turkish sentiment analysis under a unified experimental framework. Several ML models are trained on a balanced dataset of user reviews, including Linear Support Vector Machine Classifier (LSVMC), Multinomial Naïve Bayes (MNB), and Logistic Regression (LR). Their outputs were further combined using Ensemble Learning (EL) models, namely Majority Voting (MVEL) and Stacking (SEL). Results demonstrate that the SEL Classifier outperforms all examined models, achieving 92.80% accuracy at the cost of increased computational complexity. Among the examined individual models, LSVMC (92%), MNB (92%), and LR (92%) had the best accuracy. While the study does not aim to achieve state-of-the-art performance with deep or transformer-based architectures, the results demonstrate that optimized classical models remain highly effective in Turkish SA, and their accuracy can be further improved with EL mechanisms.
Ahmad Bwidani, Ali Karah Bash· Uludağ University Journal of...· 0 citations
Sentiment analysis has been applied to understand users' opinions expressed in unstructured text on various online platforms. Supervised machine learning approaches typically require large amounts of labeled data; however, annotation is expensive and laborious. This research aims to implement semi-supervised learning via pseudo-labeling to reduce reliance on labeled data and improve the performance of sentiment classification models relative to a supervised baseline. Contributions of this research include a pseudo-labeling framework implemented with limited labeled data, benchmarking of Naive Bayes and Random Forest algorithms across three Indonesian-language datasets with different linguistic characteristics (IndoNLU, E-commerce, and BCA Mobile app reviews), and an analysis of model performance sensitivity to the pseudo-labeling threshold. Count Vectorizer and TF-IDF were utilized for feature extraction with unigram and bigram parameters. Accuracy and F1-score were measured to benchmark model performance. Results show that pseudo-labeling improved performance across all classification models compared to the baseline by an average of 0.005–0.010, and up to 0.030 on the E-commerce dataset. Naive Bayes was found to have more consistent results across datasets, while Random Forest capitalized more on the implementation of pseudo-labeling under the right circumstances. Overall, TF-IDF outperformed Count Vectorizer, and larger threshold values yielded more consistent performance improvements across datasets, though lower values offered greater potential for gains on large, diverse datasets. These results suggest pseudo-labeling is a viable method for incorporating unlabeled data. Its efficacy depends on dataset attributes, feature extraction methods, and the threshold value.
Arvidion Havas Oktavian, A. Aribowo· TEPIAN· 0 citations
The paper presents a lightweight yet effective machine learning-based framework for automated sentiment classification of Hindi textual data. This proposed work addresses the persistent challenges of data sparsity, linguistic diversity, and limited annotated resources that hinder sentiment analysis in regional Indian languages. Here methodology encompasses a systematic pipeline comprising data preprocessing, Term Frequency–Inverse Document Frequency (TF-IDF) feature extraction with unigram and bigram representations, and supervised classification using Multinomial Naive Bayes (MNB) and Logistic Regression (LR) algorithms. Experiments on the IIT Patna Movie Reviews Hindi Sentiment Analysis dataset (2,480 training, 310 validation, and 310 test samples spanning the negative, neutral, and positive classes) demonstrate that Logistic Regression substantially outperforms the Naive Bayes baseline, achieving 88.19% training accuracy and 54.84% test accuracy against 67.82% and 43.23% for MNB, together with a higher micro-averaged Receiver Operating Characteristic – Area Under the Curve (ROC–AUC) (0.742 vs. 0.653). Class-wise analysis shows that positive sentiment is the easiest to detect (LR F1 = 0.63), while the neutral class remains the most challenging (LR F1 = 0.39). Proposed framework offers a computationally efficient, interpretable, and scalable solution for Hindi sentiment analysis, with direct applicability to social media monitoring, customer feedback analysis, and opinion mining in regional language ecosystems. Experimental codes is made available as a fully executable Google Colab notebook to ensure reproducibility and facilitate future research extensions.
Satyapal Singh, Jarnail Singh, D. S· International journal of com...· 0 citations
Aspect Based Sentiment Analysis (ABSA) aims to determine sentiment with respect to specific aspects of a text, providing more detailed insights than conventional sentiment analysis which assigns a single polarity to the whole text. However conventional techniques often fail to capture fine grained aspect level sentiment limiting their effectiveness in real world applications such as product reviews and customer feedback analysis. In this article proposes an innovative ABSA framework that synchronizes enhanced feature engineering with a lightweight hybrid deep learning architecture. In the proposed method Text preprocessing is done using a BERT tokenizer followed by feature extraction with an improved TF-IDF approach and Aspect Term Extraction (ATE) allowing the model to capture both global context and aspect level information. A hybrid classifier synchronizes Link-Net and SqueezeNet for fast and accurate sentiment classification. Experiment was conducted on data for Restaurant Reviews containing 10,000 reviews. The dataset splitting in training, validation and testing set 70:15:15 respectively. Fiive-fold cross-validation was applied to ensure robustness of proposed framework. Outcome of analysis demonstrate that the proposed farmwork gained a high accuracy of 97.7%, sensitivity 96.4% and F-1 Score 91.0% as compare to several traditional framework. The Proposed technique suggests a balanced tradeoff between performance and computational cost making it suitable for real time ABSA applications.
