Sentiment analysis is important in applications such as e-commerce, online opinion mining, and digital media for data-driven decision-making. Most current methods use sequential text representations and provide explainability after model training, but cannot guarantee that the relational attention learned by the model aligns with the text evidence for the target sentiment. To address this challenge, we introduce Explainability-Guided Relational Attention Regularisation (EG-RAR), which we add to the XAI-Enhanced HGTBF-OM framework. Reviews are represented as token-level co-occurrence graphs, where edges capture local token co-occurrence relationships and TF-IDF values are used to construct node-level features for relational learning. We incorporate explainability into the model optimisation process from the start, using it as a supervisory signal through the explanation-attention alignment loss and attention-entropy regularisation. We use complementary analyses such as SHAP, GNNExplainer, and attention-based analyses for post-hoc interpretation and validation. We evaluate the framework on the Zomato, Amazon, Flipkart, and IMDb datasets. The complete EG-RAR model achieves an overall accuracy of 99.31%, an F1-score of 99.30%, an AUROC of 99.45%, an average perturbation-based faithfulness gain of 14.82% and an Attention–Target Alignment of 0.713. The contribution of the proposed mechanism of learning is further supported with the help of controlled ablation experiments, randomised-relevance experiments, robustness evaluation, repeated runs, statistical testing and confidence intervals. The results indicate that EG-RAR improves predictive accuracy and relational consistency between relational attention and sentiment-relevant explanatory evidence, making sentiment analysis more transparent and trustworthy.
As one such process, opinion mining carries immense significance in deriving actionable insights from textual information, facilitating data-driven decisions across sectors. While existing methodologies have attained great strides, including hybrid deep learning models and artificial neural networks, they need help addressing deep sentiments and contextual relationships in complex datasets. The current state-of-the-art approaches either is reliant on heuristics-based methods such as LSTMs or independent graph learning techniques or less sophisticated strategies based on either one of graph-based techniques or domain transformers without ensembling their respective characteristics, thus causing them to lose the power of harmony between the two as a pioneering approach. To solve these challenges, this study presents the Hybrid Graph Transformer-Based Framework for Opinion Mining (HGTBF-OM). It integrates graph representation, TransformerConv layers, and multi-head attention mechanisms. Experimental validation on the Zomato dataset demonstrated that the proposed framework outperformed current models, achieving a commendable accuracy of 99.01%, surpassing TransLSTM, So-haTRed, and Hybrid GCN-RF. The focus on these results demonstrates the framework’s facet-wise analysis capacity and its performance on balanced data. It is helpful for e-commerce, healthcare, and social media monitoring applications where the precision of sentiment analysis is vital. This framework addresses the limitations of classical methods and provides a scalable, efficient approach to contemporary opinion-related sentiment analysis tasks. Future work plans to broaden its use on multi-lingual datasets, improve computational efficiency, and assess its usability in dynamic and task-specific sentiment analysis scenarios.
17 March 2025 in International Journal of Computational and Experimental Science and Engineering
In the contemporary era, customers' opinions are paramount as they can help improve the quality of service that leads to business growth. People from all walks of life often voice their opinions about goods and services using social media. Modern businesses cannot ignore social media feedback. As technology advances, including big data, cloud computing, and distributed computing architectures, it is now possible to analyze large volumes of data to discover trends or patterns in customer behavior. Therefore, sentiment analysis has become an important research area that helps organizations promote their businesses. Artificial intelligence, of late, is widely used to discover sentiments from customer reviews. However, analyzing sentiments accurately is nontrivial due to the complexity of procedures involved in mining textual data. There is a need to leverage performance in sentiment analysis. This research aimed to achieve this goal by proposing a DL-based framework for analyzing the sentiments of given data. We proposed a novel DL architecture, SentiNet, to efficiently classify sentiments in customer reviews. We proposed an Efficient Learning-Based Sentiment Analyzer (LBSA) algorithm, which exploits novel vectorization, embeddings, and the novel architecture of the proposed SentiNet model. Our empirical study with a benchmark dataset of customer reviews on restaurants and food items revealed that the proposed SentiNet model outperformed many existing DLs with the most accurate simulations of 98.68%. Our framework can be incorporated into business applications for sentiment analysis and improving service quality
Today, with the Evolution of technology and tools like computer systems, mobiles, tablets pieces use these http as default protocol. The fact is a huge amount of data transactions are being processed due to these the http protocol. Because of this reason http protocol is becoming a target to attackers. It is important to analyze traffic between various protocols to seek attack attempts and also take certain measures to prevent them from these attackers. In this we are going to evaluate the effectiveness of web application firewall by using Confusion matrix plots we use R studio as a tool for finding out the results. We focus on the Precision, Recall, Sensitivity, Accuracy and False Positive Rate with the help of random numbers. we used different classifiers to test the detection accuracies. The experiments show that we can also find effectiveness in a better way.