Multimodal Classification in E-Commerce: a Systematic Review

EasyChair Preprint 9469

Abstract

The 21st century is the Internet's era of big data. E-commerce has evolved swiftly, and online shopping has become prevalent. E-commerce platforms, including Amazon, and eBay are swamped with numerous product categories. These platforms must categorize the products to assist product management and recommendation but doing so manually can be highly expensive. Machine learning models that can decrease the expense and time of hiring human editors are required due to the enormous volume of new products being uploaded every day and the dynamic nature of the categories. In recent years, ML-based unimodal classification technology, such as SVM and DL, has been widely implemented in the business sector to categorize e-commerce products. Despite the positive findings published thus far, it is thought that the performance of unimodal-based algorithms can be further enhanced by incorporating multimodal product information. Thus, this paper systematically reviews and explores multimodal models used in E-commerce product classification. There are umpteen research articles published in the scientific domain from 2017 to 2022. A comprehensive literature review is missing which can help researchers and E-commerce platforms to understand and utilize the most accurate classification models.

Keyphrases: Classification, e-commerce, image classification, multimodal architecture, text classification

@booklet{EasyChair:9469,
  author    = {Karan Mehta and Yuvraj Maroo and Ria Lele and Pragati Khare},
  title     = {Multimodal Classification in E-Commerce: a Systematic Review},
  howpublished = {EasyChair Preprint 9469},
  year      = {EasyChair, 2022}}