Kateřina Macková

Large Language Models for Data Augmentation in Product Matching

These are the supplementary materials for the INLG 2026 paper: "Large Language Models for Data Augmentation in Product Matching"Abstract: Large language models (LLMs) are increasingly used for natural language processing tasks in data-scarce domains. In this paper, we study their application to product matching, the task of deciding whether two product descriptions from two different e-shops describe the same product. We use three recent LLMs (Llama~3.2, Gemma~3, and DeepSeek-R1) to enrich product records in product-matching datasets with descriptions generated from the limited available information, and we show that the generated descriptions improve the performance of both LLM-based and traditional machine-learning matchers.