Visual Analytics Course

Bordeaux roadmap / Project & Talk tentative structure

Your talk presentation and your report should adopt a structure embracing both the visualization pipeline (above), and Munzner's nested model for visualization (below).

The problem

(2-3 slides)

Although your project was primarily designed around a specific dataset, the core motivation should be questions experts are asking. You need to make these questions explicit.

The data comes next: because you want to answer these questions, you considered using this dataset hoping to find part of the answers in there. Explain why this dataset is relevant to answer the questions you have. We all are conscious here that the questions you have most probably are built around the dataset you chose first – but I want you to play the game.

The introductory part of the project document, as well as your slides, should describes the context in which those questions are asked.

The dataset

(2-3 slides)

Next comes the data. You should describe the data, as it is presented.

  • Explain how/why this data is a useful artefact to answer the questions you (or experts) are asking.
  • Explain its structure – as it is given. What raw material did you have in hand?
  • Explain how you had to process it and restructure it in order to mine it and find answers. Did you build some structure on top of the original dataset.
Your findings

(2-3 slides)

Next comes the answers you were able to formulate out of the data used to address the questions.

  • What conclusions did you reach?
  • Are they satisfacory? frustrating? remain vague? Be sharp.
  • You might want to show images or short videos of your work to support your claims.

The mining and visualization

N x (2-3 slides)

Next you will describe the work you did on the dataset and dig into technical details.

  • You'll want to save the details about the data processing phases for later. Focus on the visualization.
  • You'll go back to the questions and explain how you mine the data, or used the visualization to reach your conclusions.
    • What visualization did you build. What were the particular challenges you faced.
      • Was it about dealing with a large dataset?
      • A smaller but more complex one?
      • Was it baout combining multiple different views into a single one?
      • Was it about tuning a particular algorithm?
      • Was it about implementing an algorithm looking for particular patterns in the data?
      • Etc.

This section will typically contain several (N) subsections according to your particular problem. Maybe you only designed one visualization that required more time. Maybe you designed a few to get different perspectives on a same problem.

A live demo

(5-10 minutes)

You might want to show a live demo of a part of your work. Maybe it's worth seeing live an algorithm trying to ind a pattern in a graph. Or show live some process helping to grasp the difficulty of the problem you faced. Your presentation should definitely include some demo.

Massaging the data

(2-3 slides)

You may use some space to provide info about the preliminary phases requiring to process the data and put into a format that as convenient for the tools you used.


(2-3 slides)

This part is just as important as the preceding ones.

The first part of a conclusion is to wrap up, recall and claim that you reached your objectives, or part of them.

Why? Because most probably your answers will be incomplete, you were forced to make choices that were not optimal due to time limit. We want to know whether you are conscious of the limitations of your work.

Ok now that we've heard you were conscious your work was perfectible, we want to know what/how you would have done differently if you had more time, the data was of better quality or more reliable, etc.

  • Are the limitations due to time?
  • Poor data quality? Data too large?
  • Difficulty to deal with specific algorithms?
  • Etc.

Citing your sources

(1 slide)

It may happen that you got an idea, or part of an idea, from visiting a web site or reading a paper or book chapter. Document this in your work. Avoid citing sentences of others without giving them credit – it's way too easy to find these sentences back using Google.

/net/html/perso/melancon/Visual_Analytics_Course/data/pages/bordeaux/have_a_look_at_the_suggested_structure_for_your_talk_project_document.txt · Last modified: 2012/11/26 10:48 by melancon