abstract poetry
A generous interface for iterative and semantic search of academic papers
June-Aug 2022 (8 weeks)
︎ Deepcore AI Lab,
a Softbank Subsidiary
︎︎︎ Concept, Figma,
Github, Demo
[[]] guided-ML,
open-science,
bayesian-inference
This project demonstrates the potential for iterated semantic search to support the discovery of academic papers.
Using user impressions on paper abstracts as inputs to a Bayesian algorithm, abstract poetry traverses an embedding space of unseen abstracts to locate regions of possible interest.
Researchers found this mode of search supported “keyword discovery” better than existing modes of search , where they would begin with a fuzzy sense of what they were looking for, and end with a set of precise research keywords.
a Softbank Subsidiary
︎︎︎ Concept, Figma,
Github, Demo
[[]] guided-ML,
open-science,
bayesian-inference
This project demonstrates the potential for iterated semantic search to support the discovery of academic papers.
Using user impressions on paper abstracts as inputs to a Bayesian algorithm, abstract poetry traverses an embedding space of unseen abstracts to locate regions of possible interest.
Researchers found this mode of search supported “keyword discovery” better than existing modes of search , where they would begin with a fuzzy sense of what they were looking for, and end with a set of precise research keywords.