A semantic analysis-driven customer requirements mining method for product conceptual design.

Area of Science:

• Engineering
• Computer Science
• Human-Computer Interaction

Background:

• Accurate customer requirements are crucial for product design quality and innovation.
• Existing methods often overlook quantitative analysis in analogical reasoning environments.
• Latent customer needs can be uncovered through analogical inspiration.

Purpose of the Study:

• To propose a novel semantic analysis-driven method for mining customer requirements in product conceptual design.
• To leverage deep transfer learning and improved latent Dirichlet allocation (ILDA) for enhanced requirements acquisition.
• To address the limitations of existing approaches by incorporating analogical reasoning.

Main Methods:

• An analogy-inspired verbal protocol analysis experiment was conducted to gather elevator customer requirements.
• A deep transfer learning model using Chinese BERT (Bidirectional Encoder Representations from Transformers) was fine-tuned for requirements classification.
• Improved Latent Dirichlet Allocation (ILDA) was employed to mine functional customer requirements.

Main Results:

• The BERT deep transfer model achieved effective accuracy in classifying customer requirements into functional, behavioral, and structural domains.
• Five types of elevator customer requirements, along with keywords and their weight coefficients, were extracted using ILDA.
• The method successfully mined latent and innovative customer requirements.

Conclusions:

• The proposed semantic analysis-driven method offers a novel approach to customer requirements mining in product conceptual design.
• The integration of deep transfer learning and ILDA enhances the quantitative analysis of customer feedback.
• This research provides valuable insights for improving product innovation through natural language processing techniques.