大型語言模型建構電商專家知識系統

Constructing E-Commerce Expert Knowledge System using LLM Models

印刷電子商務能否採用大型語言模型以並建構公司內部的專家知識及報價資料為小型語言模型的服務型態。因此本研究聚焦於大型語言模型（LLM）在電子商務與企業內部專家知識系統（專業客服及專業報價）之應用，探討如何運用LLM的自然語言處理與生成能力，提升客服效率、優化知識管理流程並自動產生特定格式文件（如PDF報價單與Excel報表）。經本研究實驗結果顯示，於小型語言模型（7B）可依據公司的專家知識回答客戶的提問。並提出「餘弦相似度轉換行列資料庫解讀法」可正確從Excel的行列標纖結構找到正確的報價以即時服務客戶的詢價。

Printing e-commerce will soon adopt a large language model and build a service model that uses the company's internal expert knowledge and quotation information as a small language model. Therefore, this study focuses on the application of large language models (LLM) in e-commerce and internal expert knowledge systems (professional customer service and professional quotations), and explores how to use the natural language processing and generation capabilities of LLM to improve customer service efficiency, optimize knowledge management processes, and automatically generate specific format files (such as PDF reports and Excel quotations). The experimental results of this study show that under a small language model (7B), customers' questions can be answered based on the company's expert knowledge. We also proposed the ''cosine similarity conversion row-column database interpretation method'' to correctly find the correct quotation from the labeled row-column data structure of Excel, and provide instant service for customer inquiries.