Optimizing ECommerce Listing: LLM Based Description and Keyword Generation from Multimodal Data

Abstract

The way products are displayed on web pages is a crucial decision for e-commerce websites since it may significantly impact the sales of the products. This research explores the integration of Large Language Models with image recognition and meta-data analysis to auto-generate keywords and descriptions for e-commerce product listings. By analyzing visual content, item size, weight, color, and other meta-data, the LLM creates rich, accurate, and search-optimized product descriptions. This method aims to enhance discoverability and accuracy in e-commerce catalogs, reducing the workload for sellers and improving the shopping experience for buyers. This study proposes an advanced approach for e-commerce platforms using LLMs to generate product descriptions and keywords from multimodal data inputs, including images and meta-data. The research focuses on how the integration of visual and textual analysis by AI can create more detailed, accurate, and appealing product descriptions, ultimately leading to enhanced user engagement and sales conversion rates.