SemiChat: A Chatbot Designed Specifically for Semiconductor Tasks

Abstract

The semiconductor industry is vital for advancements in technology across various sectors, yet Large Language Models (LLMs) have been underutilized in this field due to the complexity of specialized knowledge and domain-specific terminology. To address this gap, we introduce SemiChat, a chatbot tailored for semiconductor science that leverages Retrieval Augmented Generation (RAG) technology to provide accurate, timely responses. By combining extensive domain-specific knowledge from technical literature and research with the model of qwen team, SemiChat enhances retrieval and generation capabilities, ensuring precise and contextually relevant outputs. Evaluated in five key areas of semiconductor science-advanced packaging, circuit design, gate-all-around transistors, GaN HEMTs, and hot carrier injection-SemiChat outperformed three baseline models in terms of accuracy and relevance. The experimental results demonstrate SemiChat's superiority in addressing semiconductor-related queries, offering an efficient solution for knowledge retrieval and technical problem-solving in this specialized field.

Ähnliche Arbeiten