Visualizing the performance of classification algorithms with additional re-annotated data

Authors:
Megan K. Torkildson
Affiliations:
University of Washington, Seattle, Washington, USA
Venue:
CHI '13 Extended Abstracts on Human Factors in Computing Systems
Year:
2013

Citing 4
Cited 0

A tale of two online communities: fostering collaboration and creativity in scientists and children

Proceedings of the seventh ACM conference on Creativity and cognition
Interactive optimization for steering machine classification

Proceedings of the SIGCHI Conference on Human Factors in Computing Systems
Adapting grounded theory to construct a taxonomy of affect in collaborative online chat

Proceedings of the 30th ACM international conference on Design of communication
Statistical affect detection in collaborative chat

Proceedings of the 2013 conference on Computer supported cooperative work

Quantified Score

Hi-index	0.00

Visualization

Abstract

The performance of machine learning (ML) classification algorithms in an open-ended problem with manual labels is difficult to assess, because errors can exist both in the classification and the data. This paper introduces a new visualization, confusion diamond, that exposes both kinds of errors in the context of analyzing affect in chat logs of scientists studying supernovae. I present key design elements of this visualization, relevant usage scenarios, and findings from semi-structured interviews with other members of the research team.