Transforming sentiment analysis for e-commerce product reviews: Hybrid deep learning model with an innovative term weighting and feature selection

Information Processing & Management, Journal Year: 2024, Volume and Issue: 61(3), P. 103654 - 103654

Published: Jan. 30, 2024

Language: Английский

---

Scalable deep learning framework for sentiment analysis prediction for online movie reviews

Heliyon, Journal Year: 2024, Volume and Issue: 10(10), P. e30756 - e30756

Published: May 1, 2024

Sentiment analysis has broad use in diverse real-world contexts, particularly the online movie industry and other e-commerce platforms. The main objective of our work is to examine word information order analyze content texts by exploring hidden meanings words text reviews. This study presents an enhanced method representing computationally feasible deep learning models, namely PEW-MCAB model. methodology categorizes sentiments considering full written as a unified piece. feature vector representation processed using called Positional embedding pretrained Glove Embedding Vector (PEW). these features achieved inculcating multichannel convolutional neural network (MCNN), which subsequently integrated into Attention-based Bidirectional Long Short-Term Memory (AB) experiment examines positive negative textual Four datasets were used evaluate When tested on IMDB, MR (2002), MRC (2004), (2005) datasets, (PEW-MCAB) algorithm attained accuracy rates 90.3%, 84.1%, 85.9%, 87.1%, respectively, experimental setting. implemented practical settings, proposed structure shows great deal promise for efficacy competitiveness.

Language: Английский

---

Unveiling Emotional Insights in E-Commerce: A Journey Into Visual Sentiment Analysis for User-Generated Products Through Human-Robot Interaction

SSRN Electronic Journal, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 1, 2025

In the intricate tapestry of e-commerce, where human-generated content unveils a burst sentiments within visual expressions, our research propels exploration sentiment analysis methodologies. Focused on deciphering nuanced emotional undertones user-generated content, approach integrates deep learning, semantic text analysis, and human-robot interaction. The interplay these methodologies resonates with explosion inherent in human expression, acknowledging multifaceted nature encapsulated pixels. Our methodology begins learning assisted (DLSTA), robust framework designed for emotion detection using big data. By harnessing word embeddings natural language processing, model delves into syntactic intricacies textual achieving an expressively superior rate 98.76% classification accuracy 98.67%. Expanding beyond nuances, extends to adapting developed dynamic landscape e-commerce. User-generated product images become focal points, adaptability is showcased through precision, recall, F1 score metrics across ten samples. expressions acknowledged, each image presenting unique that navigates interpretative finesse. Human-robot interaction emerges as pivotal layer methodology, injecting complexity depth analysis. between intuition computational precision mirrors capturing not only static but evolving stream encountered digital marketplace.

Language: Английский

---

Advancing sentiment analysis by addressing negation handling challenge via unsupervised mathematical approach

Social Network Analysis and Mining, Journal Year: 2025, Volume and Issue: 15(1)

Published: March 19, 2025

Language: Английский

---

Hinglish Sentiment Analysis: Deep Learning Models for Nuanced Sentiment Classification in Multilingual Digital Communication

Published: March 15, 2024

The study explores the efficiency of a hybrid LSTM-GRU deep learning model for sentiment analysis on Hinglish data, language blending Hindi and English. Integrating Long Short-Term Memory (LSTM) Gated Recurrent Unit (GRU) architectures, adeptly addresses linguistic intricacies, crucial precise classification. Leveraging combined strengths LSTM GRU, demonstrates improved memory retention accelerated training convergence, leading to superior overall performance. Impressively, achieves an accuracy 96.76%, surpassing comparable models, while precision recall scores stand at 98.49% 98.56%, respectively. Hybrid emerges as cutting-edge impactful tool in realm showcasing its promise practical deployment diverse cultural contexts.

Language: Английский

---

Predictive model for customer satisfaction analytics in E-commerce sector using machine learning and deep learning

International Journal of Information Management Data Insights, Journal Year: 2024, Volume and Issue: 4(2), P. 100295 - 100295

Published: Oct. 7, 2024

Language: Английский

---

Data preprocessing approach for machine learning-based sentiment classification

JURNAL INFOTEL, Journal Year: 2023, Volume and Issue: 15(4), P. 317 - 325

Published: Nov. 13, 2023

Public sentiment regarding a particular issue, product, activity, or organization can be measured and monitored with an application based on artificial intelligence. The data come from comments circulating social media. However, the rules for writing media have yet to standardized, so non-standard words often appear in these comments. Non-standard affect determination of into positive, negative, neutral categories. Therefore, this study proposes preprocessing approach by inserting Rabin-Karp algorithm improve words. This research consists several stages, namely crawling data, preprocessing, feature extraction, model development (based Naïve Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT) methods), analysis results. experimental results showed that proposed influences category composition. Then, testing all models obtain highest value Positive precision parameter 1. All Neutral recall parameter, almost reaching achieve f1-score average 0.95. In general, performance classification NB SVM-based better than DT method.

Language: Английский

---

BERT with an Augmented Cross-attention Decoder (BERT-ACD) for Binary and Fine-grained Multiband Sentiment Detection

Intelligenza Artificiale, Journal Year: 2024, Volume and Issue: unknown

Published: Nov. 26, 2024

In contemporary times, research in sentiment analysis has taken deeper steps into a finer and more granular analysis, transcending beyond the traditional binary or ternary classification of sentiment/opinion positive, negative, neutral. With increasing complexity challenging nature such tasks, large language models inspired by transformer architecture are frequently deployed to address challenges. Despite recorded improvements, challenges identifying different levels, strengths bands intensity aspect for which is expressed remain unresolved. this article, we propose banded system categorizing texts 7 meaningful relatable modern applications. It also capable performing aspect-based same pipeline. The model with BERT-based encoder newly proposed cross-attention, non-autoregressive decoder augmented inputs. receives an n-gram-based input sequence embedding that specifically extracted from original input, comprises list subjects, descriptive phrases, modification phrases underscore cases amplification undertone sentence. Rule-based tree parsing was use dependency extraction these inputs cross-attention decoder. Extensive experiments were conducted under setups conditions popular datasets (Amazon reviews 2023, IMDB Movies review, SST-5 SST-2 datasets) verify efficacy system. Extended labeling performed on dataset generate classes help GPT4 Bard. Experiments validate models.

Language: Английский

---

Weighted Ensemble LSTM Model with Word Embedding Attention for E-Commerce Product Recommendation

Journal of Communications Software and Systems, Journal Year: 2023, Volume and Issue: 19(4), P. 299 - 307

Published: Jan. 1, 2023

Nowadays, the proliferation of social media and e-commerce platforms is largely due to development internet technology. Additionally, consumers are used idea using these share their thoughts feelings with others through text or multimedia data. However, it difficult identify best categorization methods for this type Furthermore, users seen have difficulty understanding aspect-based conveyed by other users, currently existing models’ accuracies not up par. Deep learning models sentiment analysis (SA) provide improved performance finding out actual emotions in presented The aim research develop a weighted ensemble Long Short-Term Memory (LSTM), specialised deep model unique word embedding approaches enhance its analysis. words strong connection particular class given more weight Word Embedding Attention (WEA) technique. LSTM yields superior outcomes because excellent generalization capabilities. By integrating advantages several mitigating effects each model’s shortcomings, voting raises prediction accuracy. lessening influence outliers errors individual categorization, increases robustness categorization. This achieves 99.82 % accuracy, 99.4% precision, 99.02% f-score, 99.7% recall which much higher when compared conventional methods.

Language: Английский

---

Sentiment Analysis of Modern Chinese Literature Based on Deep Learning

Deleted Journal, Journal Year: 2024, Volume and Issue: 20(6s), P. 1565 - 1574

Published: April 29, 2024

Recent advancements in deep learning have facilitated sentiment analysis of modern Chinese literature. By leveraging techniques such as recurrent neural networks (RNNs) and transformer models like BERT, researchers can effectively gauge the expressed within literary texts. These learn intricate patterns context-specific nuances, enabling them to discern emotional tone literature accurately.Sentiment analysis, a crucial task natural language processing, plays pivotal role understanding human emotions opinions textual data. In this paper, we propose novel framework, termed BERT-LLSTM-DL, for The framework integrates Bidirectional Encoder Representations from Transformers (BERT) representation, Long Short-Term Memory (LSTM) sequential learning, feature extraction. We evaluate proposed model on dataset comprising texts achieve promising results terms accuracy, precision, recall, F1-score.

Language: Английский

---