Text-based analysis of production system documentation for cataloging historical projects using natural language processing

Efficient offer management in mechanical engineering is based on the internal availability of a central knowledge base. New customer inquiries can be analyzed and compared with historical projects. These serve as a reference with regard to technical solution concepts and calculations. Knowledge about completed projects often lies implicitly in the heads of long-term employees and is difficult to access. Building on the recent successes of language modeling using neural networks, we present a method for processing documentation of historical projects and identifying essential metadata using Name Entity Recognition (NER) for storage in a database.