Purpose: This article explores developing and integrating a product knowledge graph within an e-commerce customer support system to improve product discovery and recommendation processes. Methodology: The methodology involves a structured development process for the knowledge graph, utilizing natural language processing (NLP) to extract relevant entities from product data and machine learning algorithms to establish and categorize relationships between products. The approach integrates data from multiple sources, including vendor catalogs, online reviews, and customer interactions, ensuring a comprehensive data set. Findings: The research resulted in the creation of a dynamic, scalable knowledge graph that significantly enhances the accuracy and personalization of product recommendations. The graph’s ability to link seemingly disparate data points allows for a nuanced understanding of user behavior and preferences, improving customer satisfaction and sales performance. Practical Implications: The presented method has significant implications for retailers looking to enhance their online presence and customer interaction. By implementing this knowledge graph, retailers can expect to streamline their product recommendation processes and gain deeper insights into customer trends, which can inform broader marketing and inventory decisions. Value: This study's novelty lies in applying a comprehensive knowledge graph tailored explicitly for e-commerce systems. This graph integrates abstract and concrete entities to offer a richer, more interconnected dataset than traditional relational databases.

Bartosz Przysucha Et Al (2024). Product Knowledge Graphs - Creating a Knowledge System for Customer Support, European Research Studies Journal Volume XXVIΙ Special Issue 2, 150-159