Jul 2026· International Journal of Advanced Research in Science, Communication and Technology· pp. 417· 0 citations· 19 references
TL;DR
The findings show that although discriminative fine-tuned models continue to perform excellently in narrow classification tasks, generative LLMs possess impressive zero shot reasoning and flexibility but have issues with inconsistency and lack of transparency.
Abstract
The rapid evolution of natural language processing has driven sentiment analysis from simple lexicon based tools to sophisticated generative large language models. This systematic review charts that architectural transition, covering three distinct eras: rule based and statistical machine learning, hybrid deep learning architectures (notably CNN LSTM), and the current paradigm of transformer based and generative LLMs. Following PRISMA and Kitchenham guidelines, we synthesise findings from high impact publications (2020–2025) across IEEE, Elsevier, Springer, and ACL. The performance measures, contextual reasoning abilities, and other challenges like model variability, sarcasm detection, multimodal fusion, and interpretability form part of our analysis. Additionally, issues surrounding sustainability are addressed via Green AI techniques such as quantization and knowledge distillation. The findings show that although discriminative fine-tuned models continue to perform excellently in narrow classification tasks, generative LLMs possess impressive zero shot reasoning and flexibility but have issues with inconsistency and lack of transparency. We conclude by identifying key research gaps – deterministic benchmarking, uncertainty aware calibration, autonomous multimodal reasoning, and agentic explainability – and propose a roadmap for future work. This survey serves as a comprehensive resource for researchers and practitioners navigating the shifting landscape of sentiment analysis.
The fast development of social media has contributed to the explosion of user-generated textual content that brings about substantial sentiment analysis possibilities on a massive scale. The sentiment classification methodologies have shifted in the last ten years in terms of traditional machine learning methods to deep learning architectures and even more recent to transformer-based models. This article describes a systematic review and analytical analysis of the techniques in sentiment analysis with emphasis on deep neural and transformer-based platforms. Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, as examples of classical architectures, are discussed with respect to their structural mechanism and contextual modeling capabilities and advanced transformer models including BERT and its variants are discussed with respect to their capability to capture the complex and context-specific sentiment patterns. This study is a synthesis of reported benchmark performance on popular datasets, as opposed to new experimental simulations, to establish trends in performance, which show an increasing performance improvement between the early state of deep learning models and the current state-of-the-art transformer based models. Moreover, the paper gives a comparative analysis of the efficiency-performance trade-offs, in the aspects of computational complexity, scalability limitations, and deployment restrictions in real-time settings. Gap analysis can further be structured to detect intractable research issues such as cross domain generalization, multilingual processing, implicit sentiment detection, interpretability, and bias mitigation. This review will not only provide contextual and theoretical knowledge about sentiment analysis but also practical development tips through integrating architectural analysis, benchmark trend synthesis, and research gap identification to build robust, scalable and context-aware sentiment analysis systems in the fast changing data ecosystem.
M. Kumar, Akula Nagaraju, Gari Adarsh et al.· International Conference Com...· 0 citations
The work provides a reproducible, explainable, operationally applicable model of sentiment analysis in operationally sensitive, high-stakes Twitter sentiment analysis, and validate the hypothesis that hybrid stacking is an effective method for leveraging the complementary nature of lexical and contextual representations.
D. Abate, Nilay Mistry· International Research Journ...· 0 citations
Experimental results show that BERT-LSTM's superior performance stems from its ability to capture bidirectional contextual dependencies, improving contextual modelling of complex linguistic constructs, however, sarcasm and negation remain important sources of error.
Mohd. Danish, Saifullah Khalid· Journal of Intelligent &...· 0 citations
With the widespread growth of digital platforms, online interaction has become an essential part of everyday life. Users
frequently express their opinions, feedback, and emotions through reviews and comments on various platforms. Analyzing such
textual data plays a crucial role in understanding user sentiment and supporting effective decision-making. However, sentiment
analysis faces several challenges, including long-range dependencies within text and the presence of unknown words and
symbols. Traditional sentiment analysis approaches mainly rely on sequential models, which process text step by step and often
require higher computational time. In contrast, Transformer-based models offer improved efficiency through parallel
processing. To address these challenges, this paper presents a context-aware hybrid deep learning approach by integrating the
Robustly Optimized BERT Pretraining Approach (RoBERTa) with Bidirectional Long Short-Term Memory (BiLSTM) networks.
RoBERTa is employed to generate rich contextual word embeddings, while BiLSTM captures long-term semantic dependencies
by processing text in both forward and backward directions. The proposed model is trained and evaluated on the Twitter US
Airline Sentiment dataset comprising 14,299 samples across three sentiment classes. Experimental analysis demonstrates that
the hybrid approach achieves an accuracy of 85.14% and an F1-score of 0.8487, highlighting its effectiveness for sentiment
analysis tasks compared to baseline models
Dr. Veguru Gayatri, Dr. Rajani Rajalingam· International Journal for Re...· 0 citations
This study deploys a scalable machine learning pipeline: combining a transformer-based classifier applied to 2.01 million English-language AI-related news headlines (July 2022–July 2024) with large-language-model and human-annotator validation (three annotators, Fleiss’ κ=0.80) on stratified subsamples, to extract six interpretable, bias-linked discourse indicators computed at the AI-domain level: evaluative orientation (valence), loss salience, narrative drift, exposure-adjusted sentiment, cross-source divergence, and novelty-phase framing. Each operationalizes an established cognitive-psychology construct as a computable property of the information environment associated with biased risk–benefit reasoning. Results show systematic variation across domains: technical and methodological areas such as deep learning and natural language processing exhibit gain-salient framing, while safety-critical topics such as deepfakes (loss-to-gain headline ratio = 3.17) and facial recognition show strongly loss-salient profiles. Cross-model validation using an LLM on a stratified sample of 1000 headlines confirms that domain-level indicator rankings are robust to classifier choice (Spearman ρ=0.83; p<0.001), establishing the rank stability of pipeline outputs independently of the specific classification architecture. As a contextual application, domain-level profiles are mapped to European Union AI governance instruments, documenting parallels between discourse patterns and regulatory risk tiers. The framework provides a scalable, reproducible methodology for monitoring evaluative conditions in technology news across domains, sources, and time.
O. Topal, Inna Novalija, Joao Pita Costa et al.· Applied Informatics· 0 citations
Sentiment analysis is considered as one of the basic tasks in natural language processing (NLP) has been undergoing incredible developments with the introduction of deep learning structures. Nevertheless, traditional neural network methods are limited by major restrictions when it comes to modeling of long-range semantic interactions, contextual ambiguities, and multi-granular linguistic characteristics of affective text. In this paper, we introduce an innovative Attention based Hybrid Neural Network (ABHNN) with quantum enhanced feature representations to provide high-precision sentiment classification in a variety of textual environments. The given architecture is a synthesis of bidirectional Long Short-Term Memory (BiLSTM) networks, Transformer-based multi-head self-attention, Convolutional Neural Networks (CNNs) and quantum-inspired feature encoding layers into one end-to-end trainable pipeline. The ABHNN model proposes three main novelties, namely (i) a Quantum-Enhanced Embedding Layer (QEEL), which uses variational quantum circuits to encode the semantic features in quantum probability amplitudes; (ii) a Hierarchical Multi-Head Cross-Attention (HMCA) module, which runs simultaneously on word, phrase, sentence, levels; and (iii) a Gated Hybrid Fusion Network (GHFN) which dynamically chooses between the contribution of local convolutional feature and global recurrent-attentive feature. Massive experiments on SST-2, IMDB, SemEval-2017 Task 4, and Yelp Polarity show that the accuracy improvement is between 2.1 and 4.7% above the state-of-the-art baselines. The model has also been shown to be more robust in noisy scenarios, the domain shift, and the low-resource scenario.
Rajalaxmi Hegde, Sandeep Kumar Hegde· International Conference Com...· 0 citations