Investigating multi-label classification for bio-inspired design by using text mining and natural language processing
S. Sun · eScholarship@McGill (McGill) · 2022
Nowadays, bio-inspiration has enhanced the creation of sustainable and innovative solutions to modern engineering problems. Nature is a great source for multi-functional and optimized designs which could inspire mechanical engineers with innovative new ideas. However, it is very difficult to extract desired design knowledge from databases that are primarily text-based and focus on describing nature and biological systems. The main objective of this research is to build a multi- label classification system to classify bio-inspired designs to support design ideation. The method proposed in this study is to fuse text-based techniques such as natural language processing and text mining with machine learning to learn and predict the functionalities of bio-inspired design. Various design multi-functionalities were summarized based on the available resources from the AskNature database, then the main information extracted from the database and papers were labelled with corresponding multi-functionalities. Due to the high complexity of multi-label classification, multi-label classifiers were built based on various combinations of basic classifiers and trained to classify selected examples. One case study was conducted to verify the impact of the proposed system. The results showed that the proposed system is feasible and would be a solution for classifying the bio-inspired design and functional basis knowledge extraction method