Similarity Computation Supporting creative activities

Résumé

Creativity sessions in industry, when they are based solely on people's knowledge, produce less and less value. This is mainly due to the need to further expand the spectrum of knowledge needed to solve problems. We are therefore increasingly witnessing the limits of our knowledge capabilities to meet the demands of today's inventive problem-solving in industry. The research presented in this paper proposes a method of semantic association between problems extracted from an unstructured textual corpus of a patent using Google's Word2vec algorithm followed by cosine similarity to create original pairings between problems from different but semantically close domains. We postulate that such a method is a preamble to the automation of TRIZ and thus avoids the difficulties of not having been updated for a few decades.

BibTeX

@inproceedings{ni:hal-03078909,
  TITLE = {{Similarity Computation Supporting creative activities}},
  AUTHOR = {Ni, Xin and Samet, Ahmed and Cavallucci, Denis},
  URL = {https://hal.science/hal-03078909},
  BOOKTITLE = {{Proceedings of the Sixth International Conference on Design Creativity (ICDC 2020)}},
  ADDRESS = {Oulu, Finlande, France},
  PUBLISHER = {{The Design society}},
  PAGES = {207-214},
  YEAR = {2020},
  MONTH = Aug,
  DOI = {10.35199/ICDC.2020.26},
  KEYWORDS = {Machine Learning ; Artificial Intelligence ; Inventive Design Method ; TRIZ},
  HAL_ID = {hal-03078909},
  HAL_VERSION = {v1},
}