The results show that the proposed framework provides complementary insights: structured models support overall satisfaction prediction, while the DeBERTa-v3-based aspect model provides more detailed attribution of customer experience across product, delivery, and usability dimensions.
Abstract
Aspect-level sentiment analysis of e-commerce reviews is important for identifying fine-grained customer experience signals that are often hidden by overall star ratings. This study proposes a dual-track framework for USA-based Amazon customer reviews, combining text-driven multi-aspect sentiment classification with structured-feature satisfaction prediction. For the text-based task, reviews are modeled across three operational aspects: Product & Value, Delivery & Fulfillment, and Service & Usability, each using a three-class sentiment scheme of Negative, Neutral, and Positive. The proposed deep learning model integrates DeBERTa-v3 contextual embeddings, bidirectional recurrent refinement, and aspect-specific additive attention to separate overlapping sentiment cues within the same review. Three recurrent variants, Bi-GRU, Bi-RNN, and Bi-LSTM, are evaluated under the same modeling framework. On the held-out Amazon Customer Reviews test set, the DeBERTa-v3 + Bi-GRU + aspect-attention model achieves the best overall performance, with 0.737 accuracy, 0.757 macro-F1, and 0.760 weighted-F1, outperforming the Bi-LSTM and Bi-RNN variants. Ablation analysis further shows that sequential refinement and aspect-specific attention improve performance over encoder-only and shared-attention configurations. In the structured-feature satisfaction task, LightGBM achieves the strongest result among the evaluated machine learning models, with 82.37% accuracy and an F1-score of 0.82, outperforming XGBoost and CatBoost. The results show that the proposed framework provides complementary insights: structured models support overall satisfaction prediction, while the DeBERTa-v3-based aspect model provides more detailed attribution of customer experience across product, delivery, and usability dimensions. Overall, the study offers a reproducible and interpretable approach for fine-grained review analytics in e-commerce settings.
The rapid growth of digital platforms and online review systems has significantly increased the need for intelligent recommendation systems capable of understanding user preferences and contextual sentiments. However, traditional recommendation models often fail to capture fine-grained emotional variations, aspect-level opinions, and semantic relationships present in textual reviews, leading to reduced personalization accuracy. This work proposes a Nuanced Sentiment-Aware Recommendation with Transformers (NSAR-T) model incorporating Aspect-Based Sentiment Analysis (ABSA) with the Robustly Optimized Bidirectional Encoder Representation from Transformers (RoBERTa) to enable the generation of more granular personalized recommendations. A framework for hotel recommendations based on a combination of data sources consisting of 200,000 scraped review records from the primary source of hotels, 878,561 TripAdvisor reviews, and 37.6 million records for Expedia recommendations. The methodology consists of data preparation, extraction of aspects, sentiment classification based on RoBERTa, seven varieties of fine-grained modelling of sentiment, aggregation of sentiment, Principal Component Analysis (PCA) based feature engineering, Cross-Validation (CV), and generation of recommendations from the Top-10 list. The results show an accuracy of 99.85%, precision of 99.91%, recall of 99.81%, F1-Score of 99.86%, Root Mean Square Error (RMSE) of 0.03, Mean Average Precision (MAP) of 0.99, and Normalized Discounted Cumulative Gain (NDCG)@10 of 1; demonstrating superiority over currently employed machine learning and transformer-based recommendation systems with regard to the relevance of recommendations, contextual knowledge of recommendations, and how personalized recommendations are created.
Sonal Gupta, Indu Kashyap· 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.
Organizations can now tap into fine-grained opinions about products or service features that can be extracted from customer reviews and ratings with aspect-based sentiment analysis (ABSA). The study introduced two models, the first for aspect extraction and another one involved in sentiment analysis using bidirectional encoder representation transformer (BERT). With a testing accuracy of 98%, the aspect extraction model produces outstanding results, corroborated by metrics for precision, recall, and F1-score for every class. Furthermore, the ABSA model gives remarkable progress over earlier research, attaining an 82% testing accuracy. The proposed framework illustrates how well the BERT-based ABSA model accurately identifies and evaluates various aspects of goods or services, as indicated in customer feedback, and adds value to the body of current sentiment analysis literature, suggesting useful recommendations for improving the explanation and understanding of customer feedback. There is a chance that this research project will help consumers and businesses alike.
Arwa Akram, A. Sabir· Basrah journal of science· 0 citations
This paper introduces an efficient DistilBERT-Attention model for aspect-based sentiment analysis (ABSA), designed to balance classification accuracy against computational cost. Unlike general sentiment analysis, which assigns a single polarity to a complete review, ABSA identifies the product aspects discussed within individual sentences or clauses, such as design, quality, and price, and determines the sentiment polarity expressed toward each of them. The proposed model combines DistilBERT, a compact transformer encoder, with a modified aspect-focused attention layer that captures fine-grained sentiment signals efficiently. Experiments were conducted on 3,152 valid textual reviews, drawn from an initial collection of 3,259 Amazon India reviews of Titan watches published in 2024, across five aspect categories: design, quality, price, functionality, and comfort. The proposed model achieved an accuracy of 84.7% and an F1-score of 0.81. Compared with traditional baselines, it improved accuracy by 13.5 percentage points over a support vector machine and by 13.9 percentage points over logistic regression. Although BERT-base achieved a slightly higher accuracy of 86.1%, the proposed model retained approximately 98.4% of BERT-base accuracy while reducing memory consumption by about 40% and lowering relative processing time from 2.5x to 1.5x. The main contribution of the proposed model is therefore not absolute superiority over large transformer models, but an improved balance between accuracy, interpretability, and computational efficiency for resource-constrained ABSA applications.
Mohammad Abu Kausar, M. Nasar, Sallam O. f. Khairy et al.· Journal of Computers, Mechan...· 0 citations
In the context of large-scale online learning environments, analyzing student feedback is crucial for improving course content and learner engagement. This paper proposes MTL-BERTNet, a novel multi-task learning architecture that jointly performs aspect category classification and sentiment polarity detection from MOOC reviews. The model leverages contextual embeddings from a pre-trained BERT encoder and integrates a convolutional multi-head attention mechanism to capture subtle semantic nuances and inter-task dependencies. To further enhance shared representation learning across tasks, an inter-task matching layer (IML) is introduced. Experiments conducted on an imbalanced MOOC review dataset demonstrate strong performance, with macro F1-scores of 0.90 for aspect classification and 0.93 for sentiment prediction. These results highlight the effectiveness of jointly modeling aspects and sentiment, offering practical insights for improving course design, instructional quality, and learner satisfaction in MOOC platforms.
Raja Ouadad, Hicham Mouncif· Infocommunications journal· 0 citations
Sentiment analysis of customer reviews is an important tool in getting to know customer opinion, quality of service and product feedback. Nevertheless, classic approaches tend to label a whole review set of coarse polarity categories, which does not reflect aspect-based opinions and latent emotional trends. In order to address this deficiency, this paper proposes a new Emotion-Guided Aspect-Aware Sentiment Classification System which can effectively and reliably classify user reviews. The proposed architecture integrates the following processes: contextual preprocessing, aspect extraction, fine-grained emotion detection, memory of emotion prototype, retrieval-enhanced reasoning, emotion-sensitive feature fusion, sentiment classification and confidence-based filtering, into one architecture. The system can identify the key aspects of quality, delivery, as well as support and then emotions such as joy, anger, trust, frustration, sadness, and disappointment. These affective signals has employed in order to enhance the sentiment forecasting. Experimental data show that the proposed model reaches 93.40 percent accuracy, 92.85 percent precision, 92.30 percent recall, and 92.57 percent F1-score that are better than the base models like SVM, LSTM and BERT. The ablation experiment also verified that all the elements associated with aspect extraction, emotion detection, prototype memory, retrieval module and confidence filtering are all factors contributing to the boost in performance.
Vallem Sushma Latha, Shanker Chandre, Erukala Sudarshan· ITM Web of Conferences· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
