BERT-enhanced sentiment analysis for personalized e-commerce recommendations

Recommendation systems (RS) play a crucial role in enhancing conversion rates in e-commerce by offering personalized product recommendations based on customer preferences. However, traditional RS heavily rely on numerical ratings, which might not fully capture the subtle nuances of user preferences....

Full description

Saved in:
Bibliographic Details
Published in:Multimedia tools and applications 2023-12, Vol.83 (19), p.56463-56488
Main Authors: Karabila, Ikram, Darraz, Nossayba, EL-Ansari, Anas, Alami, Nabil, EL Mallahi, Mostafa
Format: Article
Language:English
Subjects:
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Recommendation systems (RS) play a crucial role in enhancing conversion rates in e-commerce by offering personalized product recommendations based on customer preferences. However, traditional RS heavily rely on numerical ratings, which might not fully capture the subtle nuances of user preferences. To overcome this limitation, the integration of textual data, such as reviews using sentiment analysis (SA), has gained considerable significance. Nevertheless, effectively analyzing and comprehending unstructured review data presents its own set of challenges. In this work, we propose a novel RS that synergizes collaborative filtering with sentiment analysis to deliver precise and individualized recommendations. Our approach encompasses three main steps: (1) Developing a BERT fine-tuned model for accurate sentiment classification, (2) Creating a hybrid collaborative filtering-based Recommendation Model, and (3) Improving the product selection process in the RS using BERT insights for enhanced recommendation accuracy in the e-commerce domain. Notably, our SA model exhibits remarkable accuracy, achieving 91%, and outperforming state-of-the-art models on a benchmark dataset. Through extensive experimentation and evaluation, we demonstrate that our method significantly improves the accuracy and personalization of the RS, thereby providing customers with a tailored and reliable recommendation service in the e-commerce domain.
ISSN:1573-7721
1380-7501
1573-7721
DOI:10.1007/s11042-023-17689-5