Assignment 2 Exploratory Visualization for News Finding

JOUR 479D/779D
Assignment #2 - Exploratory Visualization for News Finding

Out: October 12, 2015
Due: November, 2, 2015

Overview. This assignment is about using a standard exploratory visualization tool (Tableau) to examine and analyze a dataset for potentially interesting stories or newsworthy insights. What are the messages that we might want to communicate to a news audience? You’ll be provided a social media dataset of tweets related to Michael Brown’s shooting in Ferguson, Mo. that were collected in August 2014. In a write-up you’ll describe at least 3 interesting findings based on your analysis, including the process and screenshots of visualizations that helped you get to those findings. Also in this write-up you’ll reflect on your experience and consider any limitations, benefits, and challenges to using Tableau for this process.

Getting Started. To begin you should read Chapter 3 in Andy Kirk’s “Data Visualization: a successful design process”. This will give you a blueprint for thinking about exploring your data with an editorial eye.

The dataset that you’ll be analyzing is a subset of an archive of tweets in which people mention the word “Ferguson” in August 2014. You should read Ed Summer’s post about the archive carefully as it explains some gaps and limitations in the data that you’ll want to be aware of. The data has been further reduced in size by filtering out non-geocoded tweets.

There are two versions of the data available for you to analyze:

• One row of data for each unique tweet sent (for general analysis). CSV file.
• One row of data for each hashtag associated with a tweet (for analyzing hashtag trends). This dataset thus repeats a tweet on multiple rows if it uses multiple hashtags, because there will be one row for each unique hashtag. CSV file.

Note: When loading the data in Tableau, be sure to set the geographic role of the longitude column (“long”) correctly as “longitude”. Also be careful to examine the other auto-detected data types that Tableau assigns when loading in the data.

(You may think about combining other datasets that you find, however this is by no means essential as data acquisition could become very time consuming. You may also employ other tools for network, timeline, or text visualization as you see fit, though this is not expected and you can certainly complete the assignment with Tableau only)

Start by examining the completeness and quality of the data. Does it fit with your expectations? What is missing? Are there any errors, incomplete items, or outliers that don’t seem right? You might review the lecture on Data for other ideas of questions to ask. As you work, keep notes on what you tried, and what questions or observations you had.

Directed Analytic Navigation. Once you get a handle on what’s in the dataset start by thinking about what questions the data is capable of answering. Are there any data transformations you could do on the data (e.g. conversions, normalizations, aggregations, or other calculations) that enable new interesting analyses? What questions might have an answer that would make a headline or newsworthy finding? Even though there is no expectation that you are publishing a story from this assignment, imagine that you are exploring the data with a national news audience in mind. What would they care about?

If you are stuck coming up with questions you might consider Harcup and O’Neil’s enumeration of newsworthiness values which include: (1) reference to the elite, (2) reference to celebrities, (3) entertainment, (4) surprise, (5) bad news, (6) good news, (7) magnitude (significance), and (8) relevance to audience.

Exploratory Analytic Navigation. On the other hand you might take a more inductive approach to your analysis. Consider different ways to analyze the ranges or distributions in your data, how elements rank, the directions of trends or patterns, fluctuations, exceptions and outliers or other correlations. You might consider any of the analytic approaches that Kirk mentions on pages 67-70 of chapter 3, or that Few talks about in Chapter 4 of “Now you See it”, or that we discussed in the class on Exploratory Visualization. These may make you think of other questions that you can then pursue in a directed way as above. Try to constantly be assessing if what the image is showing you is what you would expect. What does the visualization mean in the context of the data?

Make sure to keep track of your progress by writing down questions as you go, and taking snapshots of visualizations you produced with notes about why you created the chart and what it shows.

Submission
Note: This is an individual assignment and you may NOT work in groups. All work should be your own.

Your should submit a write-up of up to ~750 words describing and contextualizing your three findings, including screenshots showing a trace of the visualizations you used and how they contributed to showing each finding, and an explanation of why you think the finding is interesting or otherwise significant. You may have to do other research to contextualize your findings as well. You do not have to create polished publishable visualizations, you just need to show your analytic process. In your write-up include a section where you reflect on your experience and consider any limitations, benefits, and challenges to using Tableau for this process.

Mail a PDF (filename of “ASGN2_<lastname>.pdf”) of your write-up to Professor Diakopoulos: nad@umd.edu before class on the due date.