Empowering Retail Dual Transformer-Based Profound Product Recommendation Using Multi-Model Review

Empowering Retail Dual Transformer-Based Profound Product Recommendation Using Multi-Model Review

Deema Mohammed Alsekait (Department of Information Technology, College of Computer and Information Sciences, Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia), Hanaa Fathi (Applied Science Research Center, Applied Science Private University, Amman, Jordan), Mohamed Taha (Faculty of Computers and Artificial Intelligence, Benha University, Benha, Egypt), Ahmed Taha (Faculty of Computers and Artificial Intelligence, Benha University, Benha, Egypt), Ayman Nabil (Faculty of Computer Science, Misr International University, Cairo, Egypt), Asif Nawaz (PMAS-Arid Agriculture University, Pakistan), Zohair Ahmed (Islamic University, Pakistan), Mohammad Alshinwan (Applied Science Private University, Jordan), Mohamed F. Issa (Faculty of Computers and Artificial Intelligence, Benha University, Benha, Egypt & University of Pannonia, Hungary), and Diaa Salama AbdElminaam (Faculty of Computers and Artificial Intelligence, Benha University, Benha, Egypt & Jadara Research Center, Jadara University, Irbid, Jordan)
Copyright: © 2024 |Pages: 25
DOI: 10.4018/JOEUC.358002
Article PDF Download
Open access articles are freely available for download

Abstract

Advancements in technology have significantly changed how we interact on social media platforms, where reviews and comments heavily influence consumer decisions. Traditionally, opinion mining has focused on textual data, overlooking the valuable insights present in customer-uploaded images—a concept we term Multus-Medium. This paper introduces a multimodal strategy for product recommendations that utilizes both text and image data. The proposed approach involves data collection, preprocessing, and sentiment analysis using Vti for images and SpanBERT for text reviews. These outputs are then fused to generate a final recommendation. The proposed model demonstrates superior performance, achieving 91.55% accuracy on the Amazon dataset and 90.89% on the Kaggle dataset. These compelling findings underscore the potential of our approach, offering a comprehensive and precise method for opinion mining in the era of social media-driven product reviews, ultimately aiding consumers in making informed purchasing decisions.
Article Preview
Top

Introduction

Electronic commerce, commonly known as e-commerce, refers to the buying and selling of goods and services via the internet. This digital evolution allows consumers to shop from the convenience of their homes or offices at any time, enhancing accessibility and flexibility. E-commerce benefits businesses by enabling them to offer round-the-clock service, which often leads to increased transactions and customer satisfaction. Platforms such as review sites, online shopping portals, and blogs empower nearly everyone to share their views on products and services (Kanwal et al., 2024). A 2024 survey by Adobe Digital Economy Index found that 85% of online shoppers now conduct extensive research online before making a purchase decision, highlighting the influential role of the internet in consumer behavior (Kumar et al., 2024).

Moreover, the emotional range expressed in online reviews—from positive to negative sentiments—can profoundly impact consumer choices and corporate reputations. These reviews not only inform potential buyers about the quality of products and services but also play a crucial role in shaping a company’s image and potentially its revenues. Positive reviews can boost consumer trust and attract more customers, while negative feedback can harm a company’s credibility and deter potential buyers (Alslaity et al., 2024). In today’s competitive market, with a plethora of options available, customers highly value the experiences shared by others. Managing and responding to customer feedback is essential for survival in the competitive realm of e-commerce. For instance, if someone is considering buying headphones, they are likely to consult online reviews to gauge the opinions of past customers, which will significantly influence their purchasing decision.

Reviews can generally be classified into two main types: quantitative, such as star ratings, and qualitative, such as lists of pros and cons, as illustrated in Figure 1. These reviews encompass annotations in the form of words or images that convey the reviewer’s sentiments. Quantitative measures, notably star ratings, are commonly used on many online retail sites to gauge consumer opinions. Classification involves organizing different types of reviews or feedback into distinct categories (Singh et al., 2024). Reviews can be emotionally categorized in various ways, including positive, negative, neutral, and these sentiments can be broken down into other specific emotions such as anger, sadness, happiness, and annoyance.

Figure 1.

Textual review example with customer ratings

JOEUC.358002.f01

Consequently, consumers increasingly rely on product reviews to gather information and make well-informed purchasing decisions (Mei et al., 2023). However, with the overwhelming amount of available information, approximately 32% of consumers report feeling confused and 30% express frustration. Despite this, many still prefer to read through several reviews before choosing a product. To mitigate these challenges, sentiment analysis (SA) has become an essential tool. It helps extract, evaluate, and present the sentiments of individuals on social media platforms.

SA, a branch of natural language processing, automates the identification of opinions from textual data (Miah et al., 2024). Its primary aim is to classify user-generated reviews as negative or positive based on the expressed sentiment regarding a specific topic. According to Liu ([date]), the terms “opinion mining“ and “SA“ are interchangeable and involve the study of behaviors, attitudes, feelings, emotions, evaluations, categorizations, opinions, and sentiments related to various aspects of services, products, or individuals.

Complete Article List

Search this Journal:
Reset
Volume 37: 1 Issue (2025)
Volume 36: 1 Issue (2024)
Volume 35: 3 Issues (2023)
Volume 34: 10 Issues (2022)
Volume 33: 6 Issues (2021)
Volume 32: 4 Issues (2020)
Volume 31: 4 Issues (2019)
Volume 30: 4 Issues (2018)
Volume 29: 4 Issues (2017)
Volume 28: 4 Issues (2016)
Volume 27: 4 Issues (2015)
Volume 26: 4 Issues (2014)
Volume 25: 4 Issues (2013)
Volume 24: 4 Issues (2012)
Volume 23: 4 Issues (2011)
Volume 22: 4 Issues (2010)
Volume 21: 4 Issues (2009)
Volume 20: 4 Issues (2008)
Volume 19: 4 Issues (2007)
Volume 18: 4 Issues (2006)
Volume 17: 4 Issues (2005)
Volume 16: 4 Issues (2004)
Volume 15: 4 Issues (2003)
Volume 14: 4 Issues (2002)
Volume 13: 4 Issues (2001)
Volume 12: 4 Issues (2000)
Volume 11: 4 Issues (1999)
Volume 10: 4 Issues (1998)
Volume 9: 4 Issues (1997)
Volume 8: 4 Issues (1996)
Volume 7: 4 Issues (1995)
Volume 6: 4 Issues (1994)
Volume 5: 4 Issues (1993)
Volume 4: 4 Issues (1992)
Volume 3: 4 Issues (1991)
Volume 2: 4 Issues (1990)
Volume 1: 3 Issues (1989)
View Complete Journal Contents Listing