Polarity of Yelp Reviews: A BERT–LSTM Comparative Study

With the rapid growth in social network comments, the need for more effective methods to classify their polarity—negative, neutral, or positive—has become essential. Sentiment analysis, powered by natural language processing, has evolved significantly with the adoption of advanced deep learning tech...

Full description

Saved in:
Bibliographic Details
Main Authors: Rachid Belaroussi, Sié Cyriac Noufe, Francis Dupin, Pierre-Olivier Vandanjon
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Big Data and Cognitive Computing
Subjects:
Online Access:https://www.mdpi.com/2504-2289/9/5/140
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850257393686937600
author Rachid Belaroussi
Sié Cyriac Noufe
Francis Dupin
Pierre-Olivier Vandanjon
author_facet Rachid Belaroussi
Sié Cyriac Noufe
Francis Dupin
Pierre-Olivier Vandanjon
author_sort Rachid Belaroussi
collection DOAJ
description With the rapid growth in social network comments, the need for more effective methods to classify their polarity—negative, neutral, or positive—has become essential. Sentiment analysis, powered by natural language processing, has evolved significantly with the adoption of advanced deep learning techniques. Long Short-Term Memory networks capture long-range dependencies in text, while transformers, with their attention mechanisms, excel at preserving contextual meaning and handling high-dimensional, semantically complex data. This study compares the performance of sentiment analysis models based on LSTM and BERT architectures using key evaluation metrics. The dataset consists of business reviews from the Yelp Open Dataset. We tested LSTM-based methods against BERT and its variants—RoBERTa, BERTweet, and DistilBERT—leveraging popular pipelines from the Hugging Face Hub. A class-by-class performance analysis is presented, revealing that more complex BERT-based models do not always guarantee superior results in the classification of Yelp reviews. Additionally, the use of bidirectionality in LSTMs does not necessarily lead to better performance. However, across a diversity of test sets, transformer models outperform traditional RNN-based models, as their generalization capability is greater than that of a simple LSTM model.
format Article
id doaj-art-3c540bedf5d04e298e31c65ab09252ca
institution OA Journals
issn 2504-2289
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Big Data and Cognitive Computing
spelling doaj-art-3c540bedf5d04e298e31c65ab09252ca2025-08-20T01:56:25ZengMDPI AGBig Data and Cognitive Computing2504-22892025-05-019514010.3390/bdcc9050140Polarity of Yelp Reviews: A BERT–LSTM Comparative StudyRachid Belaroussi0Sié Cyriac Noufe1Francis Dupin2Pierre-Olivier Vandanjon3COSYS-GRETTIA, University Gustave Eiffel, F-77447 Marne-la-Vallée, FranceCOSYS-GRETTIA, University Gustave Eiffel, F-77447 Marne-la-Vallée, FranceCOSYS-GRETTIA, University Gustave Eiffel, F-77447 Marne-la-Vallée, FranceAME-SPLOTT, University Gustave Eiffel, All. des Ponts et Chaussées, F-44340 Bouguenais, FranceWith the rapid growth in social network comments, the need for more effective methods to classify their polarity—negative, neutral, or positive—has become essential. Sentiment analysis, powered by natural language processing, has evolved significantly with the adoption of advanced deep learning techniques. Long Short-Term Memory networks capture long-range dependencies in text, while transformers, with their attention mechanisms, excel at preserving contextual meaning and handling high-dimensional, semantically complex data. This study compares the performance of sentiment analysis models based on LSTM and BERT architectures using key evaluation metrics. The dataset consists of business reviews from the Yelp Open Dataset. We tested LSTM-based methods against BERT and its variants—RoBERTa, BERTweet, and DistilBERT—leveraging popular pipelines from the Hugging Face Hub. A class-by-class performance analysis is presented, revealing that more complex BERT-based models do not always guarantee superior results in the classification of Yelp reviews. Additionally, the use of bidirectionality in LSTMs does not necessarily lead to better performance. However, across a diversity of test sets, transformer models outperform traditional RNN-based models, as their generalization capability is greater than that of a simple LSTM model.https://www.mdpi.com/2504-2289/9/5/140sentiment analysisnatural language processingLSTMBERTdeep learningpolarity detection
spellingShingle Rachid Belaroussi
Sié Cyriac Noufe
Francis Dupin
Pierre-Olivier Vandanjon
Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
Big Data and Cognitive Computing
sentiment analysis
natural language processing
LSTM
BERT
deep learning
polarity detection
title Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
title_full Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
title_fullStr Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
title_full_unstemmed Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
title_short Polarity of Yelp Reviews: A BERT–LSTM Comparative Study
title_sort polarity of yelp reviews a bert lstm comparative study
topic sentiment analysis
natural language processing
LSTM
BERT
deep learning
polarity detection
url https://www.mdpi.com/2504-2289/9/5/140
work_keys_str_mv AT rachidbelaroussi polarityofyelpreviewsabertlstmcomparativestudy
AT siecyriacnoufe polarityofyelpreviewsabertlstmcomparativestudy
AT francisdupin polarityofyelpreviewsabertlstmcomparativestudy
AT pierreoliviervandanjon polarityofyelpreviewsabertlstmcomparativestudy