Vialab member Menna El-Assady to present ‘ThreadReconstructor: Modeling Reply-Chains to Untangle Conversational Text through Visual Analytics’ at EuroVis 2018 in Brno.

On Wednesday, June 6th, 2018, a Vialab member will be presenting a new paper.

“ThreadReconstructor: Modeling Reply-Chains to Untangle Conversational Text through Visual Analytics” is lead by Ph.D. student Mennatallah El-Assady in collaboration with the University of Konstanz, and presents a visual analytics approach for detecting and analyzing the implicit conversational structure of discussions. Motivated by the need to reveal and understand single threads in online conversations and text transcripts, ThreadReconstructor combines supervised and unsupervised machine learning models to enable the exploration of generated threaded structures and the analysis of the untangled reply-chains, comparing different models and their agreement.

ThreadReconstructor will be published in the Computer Graphics Forum, volume 37, issue 3, and presented at Eurovis 2018 in Brno.

2016 Course Materials

I have made my slides, assignments, and in-class examples available for students and other instructors who may be interested. These slides are inspired by many others, in particular Dr. Mark Green who often teaches this course at UOIT. In addition, the in-class examples are adapted from a set of examples by Daniel Vogel at Waterloo, and I am grateful to him for making these available. The ray tracing assignment is based on skeleton code and an assignment handout by Dr. Tobias Isenberg. Finally, I am grateful to Dr. Aaron Hertzmann who was my graphics instructor at the University of Toronto and who set a high part for the technical content of a course like this. My course notes were valuable in preparing to teach this topic.

Course topics overview (2016):

  1. Graphics Pipeline
    1. From model to pixels, overview of the basic process
  2. Introduction to Graphics Programming
    1. GLUT, GLEW, and GLM
    2. Vertex and fragment shaders
    3. Transformations, lighting
    4. Geometrical Data
  3. Modeling
    1. Polygons, face and vertex tables, normal vectors
    2. Transformations, matrices, composition of transformations
    3. Homogeneous coordinates
    4. Implicit representations
    5. Parametric representations, piecewise representation, continuity
    6. Cubic curves, canonical form, blending functions
    7. Hermite, natural spline, Cardinal spline, Bezier curve
    8. Hierarchical modeling, OpenGL examples, display lists
    9. Subdivision algorithms
  4. Rendering
    1. Viewing transformations, projections
    2. Hidden surface, z-buffer, BSP trees
    3. Basic lighting, ambient, diffuse and specular reflection
    4. Texture mapping, Mipmaps, texture mapping in OpenGL
  5. Ray Tracing
    1. Local and global illumination
    2. Basic ray tracing technique, reflection, refraction, shadows
    3. Intersection calculations, sphere, plane, polygons
    4. Performance, bounding volumes, grids
    5. Distributed ray tracing, sampling patterns, path tracing
  6. Graphics Hardware
    1. Video, sync, frame buffers, bandwidth issues
    2. 3D acceleration, path to fixed function pipeline
    3. GPU architecture
  7. Introduction to Visualization
    1. Scientific and information visualization
    2. Visual variables and perception
    3. Colour perception & theory
    4. Colour spaces
    5. Scalar and vector visualization techniques
    6. Marching squares and marching cubes algorithms
    7. Volume rendering, transfer functions, volume traversal
  8. Advanced OpenGL programming
    1. Tessellation and geometry shaders
    2. Procedural textures
    3. GPGPU
    4. OpenGL versions
  9. Graphics Application Development
    1. Data file formats
    2. Interaction
    3. Case studies

vialab hosted talk: Dr. Nathalie Henry Riche, Microsoft Research

Nathalie Henry Riche Talk Announcement

Title: Data-driven Storytelling: Transforming Data into Visually Shared Stories

Abstract: In this talk, I will present my most recent research efforts in the field of information visualization and data-driven storytelling. While most of the research in information visualization has been focusing on designing and implementing novel interfaces and interactive techniques to enable data exploration, data visualizations also started to appear as a powerful vector for communicating information to a large audience. Stories supported by facts extracted from data analysis (e.g. data-driven storytelling) proliferate in many different forms from static infographics to dynamic and interactive applications on news media outlets. Yet, there is little research on what makes compelling visual stories or on how to empower people to build these experiences without programming. I will present insights from projects focusing on two different genres of data-driven stories: animations and comics. I will conclude the talk by reflecting on challenges and opportunities in this new research field.

Bio: Nathalie is a researcher at Microsoft Research since 2008. She is a team member of the EPIC (Extended Perception, Interaction, &Cognition) research group led by Ken Hinckley.  Nathalie holds a Ph.D. in computer science from the University of Paris Sud, France; and University of Sydney, Australia. She published her research in leading venues in Human-Computer Interaction and Information Visualization. She has received several best papers nominations and awards for her research and is involved in the organizing and program committees of major visualization and human-computer interaction conferences.

Vialab contributions to IEEE VIS 2017

Vialab members had several contributions to the IEEE VIS conference in Phoenix this month. Our contributions also represented the extent of the lab’s collaborations, from France, Scotland, Germany, Canada, and the USA.

Menna El-Assady (also affiliated with University of Konstanz) presented our paper on progressive learning of topic model parameters, for which we received an honourable mention for best paper! Her framework allows people who do not know about the inner workings of topic models to guide the settings of the parameters by examining the outputs of competing models and “voting” on their preference. Through an evolutionary approach, the topic models are refined without ever having to play with complex settings.

Hrim Mehta presented her work on Data Tours in collaboration with Dr. Fanny Chevalier and colleagues at Inria, in France. Hrim’s poster presented our idea of how to author semi-automated tours of large datasets, which can be used as a narrative overview of datasets for which a static overview would be too cluttered or overwhelming.

Mariana Shimabukuro presented her poster on automatically abbreviating text labels for visualizations. She used a crowd-sourcing platform to gather abbreviation strategies from many participants and simultaneously measured the success of these abbreviations by asking other participants to decode them. The resulting abbreviation algorithm is available as an API to abbreviate your own labels on visualizations made with d3 or other web-based platforms.

Mariana also is a co-author on an IEEE TVCG paper on font size as a data encoding, first-authored by Dr. Eric Alexander of Carleton College and colleagues at the University of Wisconsin. Eric’s talk highlighted the surprising finding that people are much better at judging differences in font size than expected, even when doing so in the presence of biasing factors such as varying length of words. This work lends credibility to the use of font size as a visual encoding, at least for tasks where “which is bigger” is the main question.

Dr. Christopher Collins was a co-organizer of the 2nd workshop on Immersive Analytics, a full-day event at VIS which attracted a number of papers and a whole lot of open research questions.

Dr. Collins, Menna El-Assady, and Dr. Adam Bradley were co-authors on Risk the Drift! Stretching Disciplinary Boundaries through Critical Collaborations between the Humanities and Visualization“, a position paper advocating for flexibility in interdisciplinary research presented at the 2nd Visualization for Digital Humanities Workshop (VIS4DH), which Dr. Collins and Menna El-Assady were also co-organizers.

Vialab member Menna El-Assady presented ‘NEREx: Named-Entity Relationship Exploration in Multi-Party Conversations’ at EuroVis 2017 in Barcelona.

We are pleased to announce that this month, a Vialab member has presented a new paper.

“NEREx: Named-Entity Relationship Exploration in Multi-Party Conversations” was lead by PhD student Mennatallah El-Assady, and presents a visualization used to explore political debates and multi-party conversations. By revealing different perspectives on multi-party conversations, NEREx gives an entry point for the analysis through high-level overviews and provides mechanisms to form and verify hypotheses through linked detail-views.

NEREx will be published in the Computer Graphics Forum, volume 36, number 3, and presented at Eurovis 2017 in Barcelona.

 

Funded PhD Position in Explainable Artificial Intelligence

Funded PhD Position in Interfaces for Explainable Artificial Intelligence

NOTE: This position is not currently available.

When an artificial intelligence system makes a decision or draws a conclusion, the reasons behind that decision are often obscure and difficult to interpret. In order to trust the outcomes of AI systems, they need to be able to present the rationale behind decisions in understandable, transparent ways. We are seeking a highly-motivated PhD candidate for an interdisciplinary research project across the fields of deep learning, visual analytics, and human-computer interaction. Specifically, the research program will be in Explainable Artificial Intelligence with a focus on creating systems to help people interpret the reasoning behind decisions made by deep learning systems. The selected candidate will join an international collaborative project and will be responsible for the design, implementation, and testing of visualization interfaces connecting to explainable machine learning systems designed by partners on the larger project.

This funded position will be established in the Visualization for Information Analysis Lab (vialab) in the Computer Science PhD program at the University of Ontario Institute of Technology in Oshawa, Ontario, Canada under the supervision of Dr. Christopher Collins. The candidate with collaborate with other team members Dr. Graham Taylor at the University of Guelph and Dr. Mohamed Amer at SRI International.

Depending on performance there is a strong likelihood of one or more paid internships at SRI International during the period of study.

REQUIRED QUALIFICATIONS/SKILLS

  • Masters degree in Computer Science, Software Engineering, Informatics/Data Science or an equivalent university-level degree and relevant experience
  • Strong and demonstrated programming skills
  • Research, work, or significant course experience in human-computer interaction, visual analytics, or interface design
  • Preference for candidates who also have experience/interest in artificial intelligence or machine learning
  • Able to work as an independent and flexible researcher in interdisciplinary teams
  • Strong English writing and speaking skills

APPLICATION PROCESS

Send the following to christopher.collins@uoit.ca:

  • Detailed CV
  • Motivation letter explaining your interest in and relevant experience for this project
  • Summary of your Master’s thesis
  • Transcripts (unofficial are acceptable at this stage; translations are not required at this stage)

Note: The selected candidate will be invited to apply through the official university application process and offers will be conditional on meeting application criteria for the UOIT CS program.

DATES & DETAILS

  • Expressions of interest as soon as possible; formal application process to follow for the invited candidate
  • Start date: September 2017 or negotiable
  • Duration: 4 years

ABOUT UOIT

The vialab at UOIT, lead by Dr. Christopher Collins, Canada Research Chair in Linguistic Information Visualization, conducts research in information analysis, visual analytics, text and document analytics, and human-computer interaction. The University of Ontario Institute of Technology (UOIT), located in Oshawa, Ontario, advances the discovery and application of knowledge through a technology-enriched learning environment and innovative programs responsive to the needs of students, and the evolving 21st-century workplace. UOIT promotes social engagement, fosters critical thinking, and integrates outcomes-based learning experiences inside and outside the classroom. Oshawa, Ontario is located near the city of Toronto, Canada, where many lab members live.

Summer and Graduate Positions at Vialab

In the summer of 2017 Dr. Christopher Collins at the Visualization for Information Analysis lab (vialab) at UOIT is seeking to hire 1 or more strong undergraduate students for research internships in a broad range of exciting topics. These internships may be eligible for credit under the UOIT co-op program.

Also, I am seeking new graduate students at the M.Sc. level; graduate studies at UOIT are funded for accepted students.

Topics include:

  • Continuing work with the UOIT Registrar’s Office on the creation of a visual analytics dashboard to analyze patterns of student grades related to a variety of factors, to improve student success and university planning.
  • Re-implementing a gesture system created to work with off-screen elements using the Leap Motion device as a polished app for deployment.
  • Developing visualizations to understand the propagation of ideas through citation networks of research papers.
  • Developing mobile applications for literacy education.
  • Investigating alternative authentication systems using subtle hand gestures / hand shakes to identify oneself to a computer system.
  • (M.Sc./Ph.D. only) Developing explanation interfaces for artificial intelligence systems (this opportunity will be linked to guaranteed industrial internships at SRI International)

Each of these positions will require students who are strong programmers. Languages used include Java, C++, Python, and Javascript. All projects will also include an aspect of literature research and report writing. Selected candidates will be expected to attend and participate in lab meetings, research discussions, and assist on other projects as needed from time-to-time. Except in exceptional cases, summer internship applicants should have completed second year in the UOIT CS program. Graduate applicants should have an Honours degree in Computer Science or related subject or be in the final year of studies toward the degree.

Interested candidates should send an email statement of interest, copy of recent unofficial transcript (mycampus print out is ok), and any other relevant materials (cv/portfolio of example projects) to Dr. Collins at christopher.collins@uoit.ca before February 28.

vialab contributions at IEEE VIS 2016

This year at the IEEE VIS Conference in Baltimore members of the lab will present papers, posters, and workshop contributions! These contributions also represent collaborations with the University of Calgary, the Univeristy of Wisconsin, Madison, and the University of Waterloo.

Papers: Optimizing Hierarchical Visualizations with the Minimum Description Length Principle and Exploring the Possibilities of Embedding Heterogeneous Data Attributes in Familiar Visualizations.

Poster: The Biasing Effect of Word Length in Font Size Encodings (InfoVis Best Poster Honorable Mention!).

Workshop papers: SpatialVis: Visualization of Spatial Gesture Interaction Logs and Visualization at LIVVIL, Digital Humanities, and the Problem of Instrumentalism at Vis4DH.

Workshop organization: Visualization for Digital Humanities.

Check out our work in Baltimore and online!

Visual Analysis of the US Presidential Debates

We are following the US presidential debates closely using the interactive visualizations created by Mennatallah El-Assady through our collaboration with the data analysis and visualization group in Konstanz, Germany. All results of the linguistic and visual analysis of the presidential debate transcripts are described on the website: http://presidential-debates.dbvis.de/
We will be updating the website regularly with new visualization findings for all upcoming debates.

 

Lexical Episode Plots

Vialab to present two EuroVis papers

We are pleased to announce that this month, Vialab members will be publishing two new papers:

“ConToVi: Multi-Party Conversation Exploration using Topic-Space Views” was lead by PhD student Mennatallah El-Assady, detailing a visual analytics approach to analyze speaker behavior patterns in conversations. It is intended to assist political science scholars in exploring the dynamics of a conversation over time using glyph-based representations.

“PhysioEx: Visual Analysis of Physiological Event Streams” was lead by PhD student Rishikesan Kamaleswaran. PhysioEx is a visual dashboard for multiple physiological data streams. By combining a Temporal Intensity Map with several other coordinated visualizations, this system is intended to assist clinical researchers in the field of neonatal medicine.

ConToVi and PhysioEx will be published in the Computer Graphics Forum, volume 35, number 3, and presented at Eurovis 2016 in Groningen.

| © Copyright vialab | Dr. Christopher Collins, Canada Research Chair in Linguistic Information Visualization |