Abstract
In the rapidly evolving landscape of e-commerce, accurately matching products to their relevant categories is crucial for improving search functionality, enhancing user experience, and driving sales. To address this challenge, this research paper proposes an innovative evaluation and annotation methodology that leverages the power of GPT (Generative Pre-trained Transformer) for product category matching. The primary goal of this study is to develop an efficient and reliable system that can automatically categorize products into appropriate groups based on their descriptions, titles, and other relevant attributes. To achieve this, we employ GPT, a state-of-the-art natural language processing model known for its proficiency in understanding and generating human-like text. The research methodology follows a multi-step approach. Firstly, a large dataset comprising product descriptions and corresponding categories is collected from diverse e-commerce platforms. Next, this dataset is carefully annotated by domain experts to establish ground truth category assignments. During the annotation process, specific challenges related to ambiguous product descriptions and overlapping categories are addressed to ensure high-quality annotations. Subsequently, the pre-trained GPT model is fine-tuned on the annotated dataset using transfer learning techniques. The fine-tuned model is then evaluated using various performance metrics, including precision, recall, F1-score, and accuracy, to quantify its effectiveness in categorizing products accurately. To validate the proposed methodology, extensive experiments are conducted on a representative set of e-commerce products. A comparative analysis is performed by benchmarking the GPT-based approach against traditional rule-based methods and other popular deep learning models in the field of text classification. The results demonstrate the superiority of the GPT-based model in product category matching, exhibiting significant improvements over existing methods. The research findings highlight the model's ability to capture complex semantic relationships between products and categories, leading to more accurate and context-aware categorization. This research paper contributes a robust evaluation and annotation methodology for product category matching in e-commerce using GPT. The study establishes the effectiveness of GPT in enhancing the performance of product categorization systems and showcases its potential in revolutionizing the e-commerce landscape. The proposed methodology holds promising implications for online retailers seeking to optimize product discovery, customer engagement, and overall user satisfaction.