Technical Term Acquisition from Production Big Data for Knowledge Sharing across LOBs and Generations

Abstract

DENSO has more than 130 production plants worldwide to produce automotive parts. A vast amount of manufacturing knowledge has been cultivated through daily production and improvement activities. However, there are limited to use because the generated useful knowledge is extremely enormous and the variety of expressions. We aim to contribute to improving productivity by developing systems that can share the valuable knowledge across LOBs and generations. As a first step, we are developing a search engine that allows inexperienced maintenance workers to easily obtain useful knowledge based on maintenance reports on the production line. For realizing this, it is necessary for experts to create a term dictionary with enormous man-hours. In this study, we verify that machine learning can be used to automatically acquire terms required for retrieval from a large amount of maintenance report documents, and it achieved 60% reduction in man-hours with better accuracy than manual term acquisition.