Abstract
Systems for organizing products on digital commerce platforms can be limited by rigid taxonomies that may not capture nuanced attributes, and some high-dimensional vector representations can lack interpretability. A computational framework can transform product information into a semantically organized and navigable space. The framework can encode products into high-dimensional vectors and then employ two parallel processes. One path may use clustering algorithms and generative language models to group products into labeled sub-archetypes. A second path may use non-linear dimensionality reduction followed by an iterative principal component analysis to extract and label continuous trade-off axes from the data. The resulting clusters and interpretable axes may be used to create user interfaces that facilitate product exploration based on continuous gradients and contextual themes, supplementing conventional categorical filters.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Grabaskas, Nathan, "Continuous Dimension Extraction for Semantic Product Organization", Technical Disclosure Commons, (September 16, 2026)
https://www.tdcommons.org/dpubs_series/11735