Bridge the gap between industrial data and Large Language Model (LLM) by mimicking the brain hemispheres function and thought process of an industrial data scientist

Industrial manufacturing generates large volumes of operational data which holds great potential for creating valuable insights to drive decision-making. However, the unique characteristics of industrial data (i.e. special terminology, code names, high accuracy requirement for quantitative analysis etc.) pose significant challenges when applying large language models (LLMs) that are predominantly trained on natural language. Instead of costly fine-tuning approach, this abstract presents an innovative approach to bridge the gap between industrial data and LLMs using technologies like named entity recognition, retrieval augmented generation and Knowledge Graphs (KG). By mimicking the brain hemispheres functions (fact-based analysis & reasoning vs. intuitive & holistic thinking) and thought process (balance and trade-off) of an industrial data scientist, we aim to optimize LLM based applications for industrial data analysis, enabling them to effectively understand and generate insights from complex datasets.