Enrico: A Dataset for Topic Modeling of Mobile UI Designs

Luis A. Leiva, Asutosh Hota, Antti Oulasvirta

Research output: Contribution to conferenceAbstractScientificpeer-review

30 Citations (Scopus)

Abstract

Topic modeling of user interfaces (UIs), also known as layout design categorization, contributes to a better understanding of the UI functionality. Starting from Rico, a large dataset of mobile UIs, we revised a random sample of 10k UIs and concluded to Enrico (shorthand of Enhanced Rico), a human-supervised high-quality dataset comprising 1460 UIs and 20 design topics. As a validation example, we train a deep learning model for three different UI representations (screenshots, wireframes, and embeddings). The screenshot representation provides the highest discriminative power (95% AUC) and a competitive accuracy of 75% (a random classifier achieves 5% accuracy in the same task). We discuss several applications that can be developed with this new public resource, including e.g. semantic UI captioning and tagging, explainable UI designs, smart tutorials, and improved design search capabilities.

Original languageEnglish
DOIs
Publication statusPublished - 10 May 2020
MoE publication typeNot Eligible
EventInternational Conference on Human-Computer Interaction with Mobile Devices and Services - Oldenburg, Germany
Duration: 5 Oct 20209 Oct 2020
Conference number: 22

Conference

ConferenceInternational Conference on Human-Computer Interaction with Mobile Devices and Services
Abbreviated titleMobileHCI
Country/TerritoryGermany
CityOldenburg
Period05/10/202009/10/2020

Keywords

  • Layout classification
  • Machine learning
  • Neural networks
  • User interface design

Fingerprint

Dive into the research topics of 'Enrico: A Dataset for Topic Modeling of Mobile UI Designs'. Together they form a unique fingerprint.

Cite this