Automatic Extraction of Design Rationale from Voice Sounds

Abstract

The Japanese manufacturing industry faces various problems, one of which is the need to solve the problem of collaboration regarding information sharing with other departments such as production engineering, manufacturing, and procurement. In this study, the target of information extraction is the idea stage in the design process. All conversations including the design rationale, such as the reasons for the decision on the structure and function of each part and the process of consideration leading up to it, are recorded in the design process, and the data is automatically transcribed using the Google Speech API. The document data is analyzed in terms of TFIDF, which is text mining technique to detect important words from a text. We proposed an automatic method for extracting design rationale and important information by extracting sentences containing the extracted words from the document data in the order of their TFIDF values, and then extracting the logical and important information sentences. The proposed method was applied to the design of scissors. The case shows the potential for extraction of rationale while there are several points to be improved.