Work

FRARBiLSTM-A Novel Fake Review Authentication Model Using Afinn and Roberta

Year: 2024
Type: article
Abstract: Due to the rise in online transactions, fake customer review identification is attracting attention. Fake customer reviews are identified using features such as reviewer identification, product information, and review text. Recent research suggests that review semantics may be particularly pertinent for text classification. The reviewers' veiled feelings could also point to misleading information. Our neural network model combines word context, customer emotions, and the traditional bag-of-words to improve fake review detection. The algorithms use N-grams, dynamic word embeddings, and emotion indicators based on lexicons to learn document-level representation. We contrast the classification performance of the detection systems with several cutting-edge methods for fake review detection to demonstrate the value of the systems. No matter the sentiment polarity or product category, the suggested approaches on the present datasets outperform Afinn, RoBERTa, Ensemble, and hybrid models. The paper offers Hybrid/Ensemble-based strategies under the proposed model named FRARBiLSTM(Fake Reviews-AFINN RoBERTa using Bidirectional LSTM). This model performs better than previous classifiers in detecting false reviews with an accuracy of 97.31. When used with Ensemble and hybrid Learning, this model can exceed and attain superior performance compared to the most modern word embedding algorithms, particularly RoBERTa and AFINN.
Language: English
Open Access status: diamond
OpenAlex ID: W7217959010
FWCI: 0.578
Cites: 42
Cited by: 3
Related to: 10