Work

Spam Reviews Detection Models in Multilingual Contexts applying Sentiment Analysis, Metaheuristics, and Advanced Word Embedding

Year: 2024
Type: dissertation
Abstract: Online reviews are a type of information that comes from consumers’ experiences after \nusing products or services. Many individuals and organizations consider these reviews \nas an evaluation source that can be used for their sake. For example, customers with the \ndesire to purchase read reviews in order to know the quality of products and services, while \norganizations view them as feedback about their products. As a result, reviews can easily \nimpact businesses’ reputations either in a positive or negative way. The COVID-19 pandemic \nled to a marked increase in the number of online reviews, with a 50% uptick observed between \n2020 and 2023. This can be attributed to the widespread adoption of remote technology and \nincreased time spent at home, as compared to pre-pandemic levels. In addition to the increase \nin numbers, the COVID-19 pandemic also led to changes in the style, structure, context, \nand preferences of online reviews. However, not all reviews are genuine, some of them are \nwritten for the purpose of misleading consumers to make a profit or damage competitors. \nSuch reviews are known as spam reviews, and they can lead to a lot of harm, including \nfinancial losses and job loss. Therefore, it is a necessity to identify these fraudulent reviews. \nSpam reviews detection is an important aspect of online security as it aims to mitigate the \npotential negative effects of such reviews, including manipulation of online reputation and \nconsumer fraud. Additionally, it helps to protect the integrity of online marketplaces and \nprevents the promotion of malicious sites through spam reviews. One of the well-recognized \nmethods used to detect spam reviews is machine learning, specifically through the application \nof Natural Language Processing (NLP). Machine learning algorithms, incorporating NLP \ntechniques, prove to be efficient methods for text classification across various tasks, such as \nsentiment analysis and spam reviews detection. In this thesis, three different approaches are \npresented in order to identify spam and the sentiment of reviews. \nThe first applied approach is the Weighted Support Vector Machine (WSVM) and a \nswarm intelligence algorithm supported by the prey and predator behaviors of soft and hard \nbesiege for spam review detection, where the optimization algorithm was incorporated for \nfeature weighting and hyperparameter optimization simultaneously. The combination of \nSVM and metaheuristic algorithms showed excellent performance in the literature for solving \ndifferent problems. To enhance their performance, a weighted method has been implemented in conjunction with a metaheuristic algorithm. The first approach is applied to test the \nperformance of the proposed WSVM in spam reviews detection and compare it with other \nexisting approaches. \nFurthermore, the second approach employed Twin SVM (TwinSVM) and a swarm \nintelligence algorithm based on the special chain movement method to address a different \nproblem, specifically in analyzing the sentiment of reviews. Due to its superior performance \ncompared to traditional SVM, particularly in mitigating overfitting, improving generalization, \nhandling imbalanced datasets, enhancing robustness, and adapting to complex structures, the \nTwinSVM model has been employed. Moreover, the swarm-based algorithm has been utilized \nin this approach for the parameter optimization of TwinSVM. TwinSVM is a newer version \nof the standard SVM, which offers several advantages compared to SVM and incorporates \nadditional parameters. Further, we used the second approach for sentiment prediction of \nthe same reviews and then compared it with the previous approach (WSVM) and other \napproaches. \nIn the final approach, we amalgamated the findings from the previous two approaches \nto pursue the optimal outcome. That is to say, sentiment analysis was applied to improve \nspam review detection based on the combination of the prey and predator behaviors of \nthe swarm-based algorithm and TwinSVM. Here, the metaheuristic algorithm is applied to \noptimize TwinSVM parameters and feature selection. In other words, in this third approach, \nwe combine the previous two applications (spam reviews detection and sentiment analysis) \nwith the optimized version of SVM (TwinSVM) and feature selection technique to enhance \nthe spam detection performance. \nGiven that there are reviews in several languages, three different languages have been \ntaken into account when extracting reviews, namely English, Spanish, and Arabic. In \norder to convert text into data that can be interpreted by machine learning algorithms, word \nrepresentation methods have been applied. Due to their ability to deal with multilingual \ntext, three pre-trained word embedding models were applied, namely, Bidirectional Encoder \nRepresentations from Transformers (BERT), FastText, and Multilingual Unsupervised and \nSupervised Embeddings (MUSE). Consequently, twenty-four datasets were generated, each \ncomprising three languages, and a multilingual dataset was created by combining these three \nlanguages with two different dimensions (rows) (100 & 400). \nThe experimental results of all approaches demonstrate the superiority of the proposed \nmethods compared with other algorithms from the State of the Art. Specifically, the prey \nand predator-based algorithm with WSVM obtained the best results (Accuracy) in the first \napproach, with 88%, 71%, 89%, and 84%, for English, Spanish, Arabic, and Multilingual \ndatasets, respectively. As for the second approach, the special chain movement algorithm combined with TwinSVM outperformed the other algorithms on most datasets. More precisely, \nit has the best results in three out of the main four datasets with 87%, 86%, and 87% \nfor Arabic, English, and Multilingual datasets, respectively. In the final approach, the prey \nand predator algorithm with TwinSVM-Fs achieved the highest accuracy when compared \nwith other algorithms, with 92%, 89%, 80%, and 85% for Arabic, English, Spanish, and \nMultilingual datasets, respectively. Further, with the help of sentiment analysis, the results \nimproved by 1.0994%, 2.6674%, 9.4430%, and 8.7448% for Arabic BERT-100, English \nMUSE-400, Spanish Fast-400, and Multi MUSE-400, respectively. At the end of each \napproach, a comprehensive analysis of the review’s style, context, and preferences was \ndiscussed. \nOverall, the thesis proposes three effective approaches to detect spam reviews, study \nthe sentiment of the reviews, and ultimately, improve the detection of spam reviews using \nsentiment analysis, all within a multilingual environment.
Language: English
Open Access status: green
OpenAlex ID: W7015467575
Cites: 283
Cited by: 10
Related to: 10