Interactive visual facets to support fluid exploratory search

Chen He, Luana Micallef, Baris Serim, Tung Thanh Vuong, Tuukka Ruotsalo, Giulio Jacucci

TL;DR

This paper introduces Interactive Visual Facets (IVF), a novel interface design for exploratory search that visualizes information facets to enhance user comprehension and control. By focusing on contextual information and rapid transitions between search criteria, IVF aims to make exploratory search more fluid and intuitive, as demonstrated through use cases and a user study.

What Problem Does This Paper Address?

Exploratory search, characterized by ill-defined goals and evolving information needs, is a common but challenging search paradigm. Traditional search interfaces, often relying on simple query boxes and ranked lists, offer limited support for the dynamic, iterative nature of exploratory search. While faceted search interfaces have shown promise by organizing information and guiding users, existing designs often struggle to provide sufficient contextual information or facilitate seamless transitions between different search criteria without disrupting the user's workflow. This paper addresses the need for interfaces that can fluidly support users as they browse, learn, and refine their search targets in complex information spaces.

Methods

The authors propose "Interactive Visual Facets" (IVF), a system that integrates a linear facet (e.g., time) and a categorical facet (e.g., keywords, contacts) with an item snippet view. The core innovation lies in two design requirements:

  1. Providing Contextual Information: IVF visualizes facets to offer an overview of the search space and the user's exploration progress. For instance, the timeline facet uses colored dots to indicate item selection status and recency, while hovering over facets can reveal relationships and summaries of related items.
  2. Supporting Rapid Transitions: A novel "filter-swipe" technique allows users to select and inspect items using facet values without immediately filtering the item space. Users can drag a categorical facet value over a linear facet bar to preview items in their intersection. This allows for tentative exploration and quick transitions between search criteria, avoiding dead ends.

Datasets

The paper uses publicly available datasets for its demonstrations and user study:

  • Enron Email Corpus: Used for email finding use cases and the user study.
  • Aweiand's Twitter Sentiment Analysis Dataset: Used for serendipitous tweet discovery.
  • cBioPortal (AML data): Used for recognizing age-related oncogene co-occurrences.

No new benchmarks were introduced; rather, existing datasets were leveraged to showcase the IVF tool's capabilities.

Key Insights and Ideas

  • Facets as Context, Not Just Filters: The most significant insight is the shift from viewing facets solely as filters to using them as dynamic contextual tools. By visualizing relationships and providing previews, facets can guide exploration without prematurely narrowing the search space.
  • "Filter-Swipe" for Tentative Exploration: The filter-swipe technique is a clever way to enable users to "try out" potential search criteria without committing. This reduces cognitive load and encourages more fluid experimentation.
  • Timeline as a Suggestion Engine: The timeline facet, when combined with dynamic suggestions, acts as a powerful context provider, helping users discover relevant leads they might not have otherwise considered.
  • Contact Suggestions are Highly Valued: In the email finding context, contact-based suggestions were found to be particularly relevant and useful for supplementing missing information, highlighting the importance of social connections in information retrieval.

Key Results

  • User Study Findings: In a user study with 11 participants performing realistic email search tasks, the IVF tool facilitated email finding.
  • Dynamic suggestions via timeline navigation were found to help users discover relevant search leads.
  • The novel "filter-swipe" design (using facets for selection without filtering) was favored over using facets as direct filters.
  • Contact suggestions were used significantly more often and found to be more relevant than keyword suggestions.
  • Qualitative Success: Three use cases (email finding, tweet discovery, and AML data analysis) demonstrated how IVF supports learning, investigation, and the formation of new search targets. The tool effectively visualized data distributions and facilitated the discovery of patterns and relationships.

Limitations and Future Work

  • Scalability: The paper acknowledges that the IVF tool's linear facet, while paginated, could benefit from semantic zooming to aggregate items and improve scalability for very large datasets. The general scalability of facets for processing high-dimensional metadata remains a challenge.
  • User Study Limitations: The user study lacked a baseline comparison, and the participant pool was limited to university affiliates, potentially hindering the generalizability of the findings to a broader population. Further studies are needed to explore the tool's effectiveness with different content types (e.g., tweets) and to generalize findings.
  • Filter-Swipe Effort: The filter-swipe technique requires continuous mouse depression, which can be effortful for detailed item inspection.
  • Future Work:
    • Implement semantic zooming for the linear facet.
  • Enhance item snippet views with highlighting of relevant keywords.
  • Improve visual encoding of item selection order (e.g., using position in addition to saturation).
  • Conduct user studies with baselines and more diverse participants to validate findings and explore generalizability.

Summary updated

Abstract

Exploratory search starts with ill-defined goals and involves browsing, learning, and formulating new targets for search. To fluidly support such dynamic search behaviours, we focus on devising interactive visual facets (IVF), visualising information facets to support user comprehension and control of the information space. We reviewed existing faceted search interfaces and derived two design requirements (DR) that have not been fully addressed to support fluid interactions in exploratory search. We then exemplified the requirements through devising an IVF tool, which coordinates a linear and a categorical facet representing the distribution and summarisation of items, respectively, and providing context for faceted exploration (DR1). To support rapid transitions between search criteria (DR2), the tool introduces a novel design concept of using facets to select items without filtering the item space. Particularly, we propose a filter-swipe technique that enables users to drag a categorical facet value sequentially over linear facet bars to view the items in the intersection of the two facets along with the categorical facet dynamically summarising the items in the intersection. Three applications with various datasets demonstrate how the features support information discovery with ease. A comparison with a baseline system provided evidence that the task performance of the IVF tool was comparable to the typical query search interface. Another study of 11 participants with realistic email search tasks shows that dynamic suggestions through the timeline navigation can help discover useful suggestions for search; the novel design concept was favoured over using facet values as filters. Based on these practices, we derive IVF design implications for fluid, exploratory searches.
BibTeX
@article{he2021interactive,
  author = {Chen He and Luana Micallef and Baris Serim and Tung Thanh Vuong and Tuukka Ruotsalo and Giulio Jacucci},
  title = {{Interactive visual facets to support fluid exploratory search}},
  journal = {J. Vis.},
  year = {2021},
  doi = {10.1145/3481549.3481565},
}
Figure