Natural Language Processing for Supply Chain with Chat GPT

We researched how an aircraft manufacturer found keywords for component specifications in the order document. The component specifications in the document were unstructured and manually reviewed by employees. This often caused problems such as overtime work and missing several specifications. Through a literature review using GPT-3, we extracted keywords of component specifications. We established a GPT-3 model for keyword extraction. First, we created training samples with prompts and responses. Then, this training sample was used to create fine-tuned models. The results showed the high accuracy of the model and the possibility of GPT-3 for the keyword extraction.