State of The Art Report (STAR) and Survey

As part of WP3 we will be delivering a STAR. The target venue is Eurovis, as they have a submission category for STAR reports. The deadline for the submission of complete STARs for Eurovis 2018 is 23:59 UTC/GMT, October 1 2017. This is different to previous years, as it appears from the call for papers that the star need to be complete, not just a short version. It is also possible to submit a STAR directly to Computer Graphics Forum. The length of the STAR can be up to 20 pages, excluding references.

As a complete survey is a key element of our STAR, we will be defining a methodology by which the survey is conducted. It is clear that we will need to do a complete search within the primary info vis and graph vis conference and journals. However, as our work features significant involvement from application domains, it would be wise to more beyond this scope and look in journal and conferences from these domains.

Keyword Definition

It is often noted that that nomenclature for Multilayer graphs is still not fully stable. As a result we have a significant list of terms which may be used to describe graph types which fit in the Multilayer framework.

When searching a publication for papers related to Multilayer graph and visualization the following terms should be used

  • Multilayer ( and variants such as multi-layer)
  • Visualization ( and visualisation)
  • Heterogeneous
  • Multilevel
  • Multirelational
  • Multimodal
  • Faceted
  • Multiplex
  • Multilabel

Infovis Domain

Given our exiting knowledge of the infovis domain it is straightforward enough to to a complete survey of the relevant publications. The identified venues so far are:

  • IEEE TVCG (inc. Infovis and Vast )
  • CHI /SIGCHI/TOCHI
  • CGF/Eurographics
  • AVI & PacificVis
  • Graph Drawing and Network Visualization (Formerly Graph Drawing)
  • Information Visualization (Sage)

Application Domains

It will not be possible to do such a complete survey within our application domains for two reasons.

  1. We have a large number of application domains
  2. We do not even have a list of specific journals from these domains.

Many Multilayer networks visualizations can appear in the application domains, so we will need to look there as well. To do to a full methodological review of these domains could be very time-consuming, and impossible to be sure of. We also have to answer the question of how do we determine which domain journals are relevant. Our approach for our application domains will be to search via google scholar, as well as list techniques we find in our initial infovis search. With that in mind, while we should be aware of which journals are relevant for visualization in the domains. Application Domains:

  1. Bioinformatics / Biology
  2. Digital Humanities
  3. Sociology
  4. Mathematics / physics / Complex systems
  5. TBC

Potential Journals of interest:

  1. BMC Bioinformatics ( Biological pathways can contain multiple layers)
  2. TBC

Example Queries

The following queries were used to query different sources and venues (Ieee explore, . Due to the different mechanisms used by various online repositories, and differences in syntax, the queries between sites differ slightly. For example, for CGF we limit search terms to the abstract, as a full search using the terms returned a vast number of documents. The key goal of searching in this manner is to have a reproducible method of finding papers.

IEEE Explore

IEEE TVCG

(“Publication Title”:PacificVis) AND (multilevel OR multi-level OR multilayer OR multi-layer OR heterogeneous OR faceted OR multirelational OR multi-relational OR multimodal OR multi-modal OR multidimensional OR multi-dimensional OR multiplex OR multilabel) AND (visualization OR visualisation) AND (graph OR network)

Pacific Vis proceedings

(“Publication Title”:PacificVis) AND (multilevel OR multi-level OR multilayer OR multi-layer OR heterogeneous OR faceted OR multirelational OR multi-relational OR multimodal OR multi-modal OR multidimensional OR multi-dimensional OR multiplex) AND (visualization OR visualisation) AND (graph OR network)

ACM Portal

AVI proceedings

(multilevel “multi-level” multilayer “multi-layer” heterogeneous faceted multirelational “multi-relational” multimodal “multi-modal” multidimensional “multi-dimensional” multiplex multilabel) AND (acmdlTitle:(“Advanced visual interfaces”) AND (visualization visualisation) AND (graph network)

CHI proceedings (multilevel “multi-level” multilayer “multi-layer” heterogeneous faceted multirelational “multi-relational” multimodal “multi-modal” multidimensional “multi-dimensional” multiplex multilabel “multi-label”) AND acmdlTitle:(+“Conference on Human Factors in Computing Systems”) AND (visualization visualisation) AND (graph network)

Transaction on CHI (multilevel “multi-level” multilayer “multi-layer” heterogeneous faceted multirelational “multi-relational” multimodal “multi-modal” multidimensional “multi-dimensional” multiplex multilabel “multi-label”) AND acmdlTitle:(+“ACM Transactions on Computer-Human Interaction”) AND (visualization visualisation) AND (graph network)

Wiley

Computer Graphics Forum

Computer Graphics Forum in Publication Titles AND (multilevel OR multi-level OR multilayer OR multi-layer OR heterogeneous OR faceted OR multirelational OR multi-relational OR multimodal OR multi-modal OR multidimensional OR multi-dimensional OR Multiplex) in Abstract AND (graph OR network) in Abstract AND (visualization OR visualisation) in All Fields

Springer

Graph Drawing / lecture notes in computer science

There was no clear way to limit the search to the grpah drawing subset of LNCS so graph drawing was added as a search term. There were a large number of results, (approx 300). When manually searched through there were no results which seemed probably related to our survey.

(“Graph Drawing”) AND (multilevel OR “multi-level” OR multilayer OR “multi-layer” OR heterogeneous OR faceted OR multirelational OR “multi-relational” OR multimodal OR “multi-modal” OR multidimensional OR “multidimensional” OR multiplex OR multilabel OR “multi-label”) AND (visualization OR visualisation)

Sage

Information Visualization

for [[Abstract multilevel] OR [Abstract multi-level] OR [Abstract multilayer] OR [Abstract multi-layer] OR [Abstract heterogeneous] OR [Abstract faceted] OR [Abstract multirelational] OR [Abstract multi-relational] OR [Abstract multimodal] OR [Abstract multi-modal] OR [Abstract multidimensional] OR [Abstract multi-dimensional] OR [Abstract multiplex] OR [Abstract multilabel]] AND [[Abstract network] OR [Abstract graph]] AND [Abstract network graph] within Information Visualization

The survey spreadsheet will be stored on google docs and is available here..

The tab “Visualization.initial.list” contains a starting list of papers. This list is a work in progress and has been compiled so far based on the queries above. Due to the broad nature of some of the queries there are many documents which may not be related to our survey. So far no documents have been added this list from the application domains. The proposed approach is, once this list is considered complete, papers will be assigned to project team members. All of the papers that are relevant to multilayer graphs, should be added to the Survey Analysis tab.

Within the survey spreadsheet text fields describing visualization text etc should be kept short ( a few hundred words). The survey spreadsheet Analysis tab will contain the following columns

  • Title: The full title of the paper
  • Authors: The authors of the paper (First-name Last name and semicoln separated between authors)
  • URL: A link to an online version of the paper if available
  • Venue: Name of the conference or journal
  • Year: Year of publication
  • Domain: domain of the application being used to demonstrate (select from drop down)
  • Nomenclature: The name used in the paper for the graph type, e.g. Multiedge, multiplex, heterogeneous, multilayer
  • Data Description: What data sets are used in the paper for demonstration / Evaluation examples?
  • Layers: the definition of layers short semicolon separated descriptions of each layer e.g. institutions;people;documents, proteomics;metabolomics, phone-gps;road-network
  • Layer Type: How the layer is defined ? e.g. categorical node attribute, attribute range, temporal attribute, edge categorical attribute, etc.
  • Graph Vis Type: chose form drop down (Node-Link, Matrix, Hybrid, Multiple Types, Other)
  • Novel Graph Vis: true if the authors approach to graph vis is not seen elsewhere
  • Graph Vis Description: Short description of key characteristics of the graph vis
  • Novel Interaction: true if the the multilayer interaction techniques are not seen elsewhere.
  • Interaction Description:Short description of key characteristics of the interaction
  • Attribute Visualization: True if the paper includes looking at the attribute data in the multilayer graph
  • Attribute Visualization Description: Short description of key characteristics of the interaction
  • Multilayer Tasks: True if the paper includes a task the may be included in a multilayer task taxonomy
  • Task Description: Short description of key characteristics of tasks
  • Novel Multilayer Analytics: True if the paper includes novel multilayer analytics
  • Multilayer Analytics Description: A short description of the analytic techniques used
  • Empirical Evaluation: true / false whether there was an Empirical (experimental) evaluations
  • Evaluation Description: A short description of the evaluation done, empirical or otherwise
  • Strengths: A description of the the strengths of the paper, a few hundred words max
  • Weaknesses: A description of the the weakness of the paper, a few hundred words max
  • Reviewer Name: Which BLIZAAR team member performed the review
  • Reviewer Comments: Other comments the reviewer would like to add (optional)
  • Survey Relevance Score: On a scale of 0 to 5, how relevant the paper is to our survey. 0 implies nothing to do with Multilayer graphs, 3 means quite relevant and we will most likely include it, 5 implies that it is absolutely vital

Assignment of responsibilities for survey

TBC: Who is responsible for reviewing papers for the survey from each of the sources.

STAR Structure

The proposed structure of the star is as below. Responsibility for the various sections will be assignedf once the survey of the papers is complete.

Introduce concepts, give simple example and emphsise the improtance of complex networks and multilayer graphs as a domain

help make the case of why this paper is needed, in addition to Previous STARs (be sure to distinguish from faceted visualization). Clarify the nomenclature as well.

Describe our approach to the survey help make the case of why this paper is needed, in addition to Previous STARs (be sure to distinguish from faceted visualization). Clarify the nomenclature as well.

This section will describe the results of the survey under various subheadings.

Application Domains of Multilayer graphs

Show importance of topic, by demonstrating the large number if application domains.

Layer Definitions

Describe various layer definitions encountered in the survey, comment on trends etc.

Visualization Approaches

Describe various Visualization approaches in the survey, comment on trends/ shortcomings etc.

Interaction Approaches

Describe various Visualization approaches in the survey, comment on trends/ shortcomings etc.

Attributes and Analytics

Discuss the treatment of multivariate data as well as new analytics that can be applied to multilayer graphs. Look at the multilayer analytics described by Kivela et al, amongst others.

Tasks for Multilayer Graphs

Describe tasks for Multilayer graphs, emphasise those not covered by existing tasks taxonomies

Opportunities in visualization & Interaction

What potential techniques are currently under-explored?

Multilayer Graph Task Taxonomy

Describing tasks that are not covered by existing taxonomies, or contextualizing tasks into the existing taxonomies where there is overlap.

New analytics, domains and solutions

Consider new domains where multilayer visual analytics may be helpful. Discuss whether using the multilayer analytics and metrics provided by Kivela et al, and others, address previously unanswered problems.

Outstanding challenges

Sketch out future directions for research in the field.

Work discussion

Scope. What will be our main focus for the paper (is it the same as one of those more a more recent survey)? focusing on application domains? should we try to cover all of these topics but at some high level or in deep in all sections.

Sumarizing. How should we list the meta-survey data e.g. one figure or table summarizing but also organizing the spreadsheet results? Should this be applied for each section individually (e.g. vis, interaction, task taxonomies) and/or overall.

Methodology. Which time periods of the journal issues should be reviewed e.g. 1995-2017?, indicate if we further checked down referenced papers (how many levels), minimum number of citations? We should specify the queries used in more detail (so it can be repeatable). Should we include a section specifically on empirical evaluations conducted?

Survey. Some of the previous state-of-the-art papers conducted a survey to experts to ask them to suggest another paper could have been missed otherwise, or to suggest challenges in the field (the latter may be interesting?).

Tool to explore the papers ? Should we provide (or use) a tool to explore these papers as done in e.g. F. Beck, M. Burch, and S. Diehl, “A taxonomy and survey of dynamic graph visualization,” Computer Graphics …, 2016. (see http://dynamicgraphs.fbeck.com). (They used tags to classify papers in this case.)

Work. For paper writing overleaf will facilitate the process I believe (and includes a git repository). As we review papers, and complete the spreadsheet, we may restructure. In the spreadsheet, we may also review state-of-the-art papers themselves, which would need to referenced in our star, but may also provide relevant paper listing to start the work.

Some Survey papers & books:

Read more at http://www.cambridge.org/fr/academic/subjects/computer-science/computing-and-society/multilayer-social-networks#44hTBYjBKtZsj4X8.99