Vol. 55 No. 4 (2018)   New Power Domain Projects
Technical Papers

Creating New Value using Mass Text Data

Text Mining for Performance Improvement

KAZUKI OZAKI
SHUN YAMAGATA
HISASHI NISHIKI
TAKASHI TAGUCHI

Text data, such as incidents and their causes relating to incompatibilities that occur during manufacturing, troubleshooting during facility operation, patent information, etc., is stored in general as unstructured data, unlike numerical data which expresses physical quantities, etc., in table form that can be structured. Therefore, the development of text data analysis has been quite slow, hindering its utilization. If such text data could be quantified and structured, classifying sentences and searching for similar ones would be possible by utilizing an analytical approach to numerical data, from which the effective use of text data such as a quick response to newly occurring incidents could be expected. This paper introduces text mining technology and application examples where text data could be quantified and structured.