mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling

Mohammad Chegini, Jürgen Bernard, Lin Shao, Alexei Sourin, Keith Andrews, Tobias Schreck

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Abstract

Many machine learning algorithms require a labelled training dataset. The task of labelling a multivariate dataset can be tedious, but can be supported by systems combining interactive visualisation and machine learning techniques into a single interface. mVis is such a system, providing a unified ecosystem to explore multivariate datasets and execute machine learning algorithms to build labelled datasets.
This paper describes a pre-study evaluation of the mVis system, comprising case studies in two different domains: collaborative intelligence and daily activities. In each case study, a volunteer researcher was asked to use mVis to explore, analyse, and label their own dataset in their own environment, while thinking out loud. The case studies provided valuable leanings in terms of the usability of the system, understanding how different analysts work, and identifying important missing features.
Original languageEnglish
Title of host publicationProceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems
Number of pages4
Publication statusAccepted/In press - 21 Oct 2019
EventIEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems - VANCOUVER, BC, CANADA, Vancouver, Canada
Duration: 21 Oct 201921 Oct 2019
https://eviva-ml.github.io/

Workshop

WorkshopIEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems
Abbreviated titleEVIVA-ML
CountryCanada
CityVancouver
Period21/10/1921/10/19
Internet address

Fingerprint

Labeling
Learning systems
Learning algorithms
Ecosystems
Labels
Visualization

Keywords

  • Visualisation
  • Machine Learning

Cite this

Chegini, M., Bernard, J., Shao, L., Sourin, A., Andrews, K., & Schreck, T. (Accepted/In press). mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling. In Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems

mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling. / Chegini, Mohammad; Bernard, Jürgen; Shao, Lin; Sourin, Alexei; Andrews, Keith; Schreck, Tobias.

Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems. 2019.

Research output: Chapter in Book/Report/Conference proceedingConference contributionResearchpeer-review

Chegini, M, Bernard, J, Shao, L, Sourin, A, Andrews, K & Schreck, T 2019, mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling. in Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems. IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems, Vancouver, Canada, 21/10/19.
Chegini M, Bernard J, Shao L, Sourin A, Andrews K, Schreck T. mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling. In Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems. 2019
Chegini, Mohammad ; Bernard, Jürgen ; Shao, Lin ; Sourin, Alexei ; Andrews, Keith ; Schreck, Tobias. / mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling. Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems. 2019.
@inproceedings{540b4a2346354c6d905c9b9fa71a9b51,
title = "mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling",
abstract = "Many machine learning algorithms require a labelled training dataset. The task of labelling a multivariate dataset can be tedious, but can be supported by systems combining interactive visualisation and machine learning techniques into a single interface. mVis is such a system, providing a unified ecosystem to explore multivariate datasets and execute machine learning algorithms to build labelled datasets.This paper describes a pre-study evaluation of the mVis system, comprising case studies in two different domains: collaborative intelligence and daily activities. In each case study, a volunteer researcher was asked to use mVis to explore, analyse, and label their own dataset in their own environment, while thinking out loud. The case studies provided valuable leanings in terms of the usability of the system, understanding how different analysts work, and identifying important missing features.",
keywords = "Visualisation, Machine Learning",
author = "Mohammad Chegini and J{\"u}rgen Bernard and Lin Shao and Alexei Sourin and Keith Andrews and Tobias Schreck",
year = "2019",
month = "10",
day = "21",
language = "English",
booktitle = "Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems",

}

TY - GEN

T1 - mVis in the Wild: Pre-Study of an Interactive Visual Machine Learning System for Labelling

AU - Chegini, Mohammad

AU - Bernard, Jürgen

AU - Shao, Lin

AU - Sourin, Alexei

AU - Andrews, Keith

AU - Schreck, Tobias

PY - 2019/10/21

Y1 - 2019/10/21

N2 - Many machine learning algorithms require a labelled training dataset. The task of labelling a multivariate dataset can be tedious, but can be supported by systems combining interactive visualisation and machine learning techniques into a single interface. mVis is such a system, providing a unified ecosystem to explore multivariate datasets and execute machine learning algorithms to build labelled datasets.This paper describes a pre-study evaluation of the mVis system, comprising case studies in two different domains: collaborative intelligence and daily activities. In each case study, a volunteer researcher was asked to use mVis to explore, analyse, and label their own dataset in their own environment, while thinking out loud. The case studies provided valuable leanings in terms of the usability of the system, understanding how different analysts work, and identifying important missing features.

AB - Many machine learning algorithms require a labelled training dataset. The task of labelling a multivariate dataset can be tedious, but can be supported by systems combining interactive visualisation and machine learning techniques into a single interface. mVis is such a system, providing a unified ecosystem to explore multivariate datasets and execute machine learning algorithms to build labelled datasets.This paper describes a pre-study evaluation of the mVis system, comprising case studies in two different domains: collaborative intelligence and daily activities. In each case study, a volunteer researcher was asked to use mVis to explore, analyse, and label their own dataset in their own environment, while thinking out loud. The case studies provided valuable leanings in terms of the usability of the system, understanding how different analysts work, and identifying important missing features.

KW - Visualisation

KW - Machine Learning

M3 - Conference contribution

BT - Proceeding of IEEE VIS 2019 Workshop on Evaluation of Interactive Visual Machine Learning Systems

ER -