Similarity metrics vs human judgment of similarity for binary data: Which is best to predict typicality?
Similarity measures for binary data have been subject to a number of comparative studies. In contrast to these studies, we provide a comparison of similarity measures with human judgment of similarity. For this purpose, we utilize the phenomenon of typicality, whose definition is based on similarity...
Saved in:
Published in | Applied soft computing Vol. 153; p. 111270 |
---|---|
Main Authors | , |
Format | Journal Article |
Language | English |
Published |
Elsevier B.V
01.03.2024
|
Subjects | |
Online Access | Get full text |
ISSN | 1568-4946 1872-9681 |
DOI | 10.1016/j.asoc.2024.111270 |
Cover
Summary: | Similarity measures for binary data have been subject to a number of comparative studies. In contrast to these studies, we provide a comparison of similarity measures with human judgment of similarity. For this purpose, we utilize the phenomenon of typicality, whose definition is based on similarity. We observe how well the similarity of objects – either computed by a similarity measure or provided by human judgment – enables the prediction of typicality of these objects in various human categories. In doing so, we examine a large variety of existing similarity measures, and utilize recently available extensive data involving binary data as well as data on human judgment of similarity and typicality.
•Comparison of 69 existing similarity measures for binary data is provided.•Human judgment of similarity is also included in the comparison.•Ability of similarities to predict typicality is assessed.•Best similarity measures are identified for prediction of typicality.•Extensive psychological data is utilized enabling the study. |
---|---|
ISSN: | 1568-4946 1872-9681 |
DOI: | 10.1016/j.asoc.2024.111270 |