Measures for Extracting Basic-Level Categories
Mariusz Mulka, Wojciech A. Lorkiewicz · 2018
This paper focuses on cognitive computing approaches to establishing basic-level categories. In particular, it provides a formal overview of four classical concepts (originating from psycholinguistic research)feature-possession, category utility, category attentional slip, and category's cue validity. Additionally, two new measures enriching the latter concept are introduced: category's cue validity with global threshold and category's cue validity with feature possession. Finally, all models are tested against 14 classical datasets (originating from psycholinguistic research)with an underlying three level category structure. As such, performed experiments allow for establishing baseline performance and basic properties of different approaches. Obtained results indicate that two new additions perform on par with the feature-possession, category utility, and category attentional slip approaches.