Supervised Process of Un-structured Data Analysis for Knowledge Chaining

Along the product life-cycle, industrial processes generate massive digital assets containing precious information. Besides structured databases, written reports hold unstructured information hardly exploitable due to the lack of vocabulary and syntax standardization. In this paper we present a methodology and natural language processing approach to exploit these documents. Our method consists in providing connections based on supervised retrieval of domain-specific expressions. No prior document analysis are required to drive the algorithm. It underlines a scale of specificity in pattern visualization. This allows relevant and specific information extraction for feedback (e.g. design stage, after-sales service).