Just popped up in my Twitter, and seemed relevant to us!
Monthly Archives: April 2016
What does Hillary Clinton’s Inbox look like?
I, too, am tired of hearing about Hillary’s use of a private email server. On the other hand, it led to a pretty neat data set to unpack: a dump of emails she’s sent and received.
I played around with this data set a bit and was particularly interested in how different groups of people interacted with Hillary. Did men use shorter sentences than women, for example? Did her staffers send one-liners versus ambassadors who sent lengthy emails? Did she have interesting relationships with people we might not be familiar with?
I didn’t get a chance to answer all of these questions, but I ended up being interested in the way words in her email were clustered, and decided to come up with a visualization based on that.
For a simple representation to start, I created a scatter plot visualization using mpld3, which creates interactive matplotlib graphs for the browser. It’s clunky to navigate (you need to switch to a zoom-in mode, drag a rectangular portion of the graph to zoom in on, then switch again to the cursor mode to scroll over words), but it’s interesting to see which words appear together for a first step.
Isn’t it interesting that “bipartisan” appears well outside the main cluster of words?
Lesson learned along the way: visualizing text is hard. I found that the norm for text visualizations out there, such as word clouds or circle packing, was reductionist for some of the data I have, like topic models or k-means clustering.
While I didn’t create data visualizations for some of the questions I posed earlier, I do have some statistics:
For males:
6187 sentences
83764 word tokens
10762 word types
7.78 average tokens per type
13.54 average sentence length
5.01 average word token length
7.34 average word type length
Hapax legomena (words that appear only once – an indicator of vocabulary usage) comprise 49.60% of the types
For females:
22845 sentences
369517 word tokens
30386 word types
12.16 average tokens per type
16.17 average sentence length
4.94 average word token length
7.84 average word type length
Hapax legomena comprise 49.76% of the types
I analyzed all of Taylor Swift’s lyrics so you don’t have to.
At the 58th Grammy Awards earlier this year, Taylor Swift became the first woman to win Album of the Year twice for a solo album.
By the numbers, this shouldn’t come as a shock. Swift — an objectively gifted singer, songwriter, and performer — has had a wildly successful career by any metric. That said, if I had to list the top 10 female performers of my lifetime I’m not sure Swift would make the cut. As culture critic Camille Paglia so delicately put it for The Hollywood Reporter, I find her music to be “mainly complaints about boyfriends, faceless louts who blur in her mind as well as ours.”
While the internet is rife with Taylor Swift listicles analyzing the lyrics of her songs, data-driven analysis is scarce (or, more likely, just private). So, in the spirit of collect and verify, I decided to do a textual analysis of TSwift’s work using Word Counter to see just how boy-centric her lyrics actually were.
True to Sands prediction from last class: 80% of my time was spent on data collection, 15% was spent sifting through said data, and I’m wrapping up the remaining 5% now. Using the database AZLyrics, I combed through the many, many songs of Taylor Swift. To date, she has released five studio albums, two live albums, two video albums, two extended plays (EPs), 37 singles, three featured singles, and eleven promotional singles To keep things simple, I decided to stick with her five studio albums, Taylor Swift (2006), Fearless (2008), Speak Now (2010), Red (2012), and 1989 (2014).
Word Counter is a pretty straightforward tool: it counts the words, bigrams, and trigrams in a plain text document which you can either paste directly into the browser or upload to the site. From there, you can download the single word counts, bigrams (2 contiguous words), and trigrams (3 contiguous words) into .csv format. Between the five albums, I copied in text from 69 songs and then downloaded the data.
Then the process became a bit less straightforward. Comparing single word-counts of individual songs and albums side by side didn’t really give me a ton of useful insight — not to mention, it’s a fairly boring way to see the data. I decided to compare Swift’s two “Albums of the Year” — 1989 (in blue) and Fearless (magenta) — by plugging the songs’ text into Tagul, a very user friendly word cloud art generator.
Other than showing Ms. Swift is a thematically consistent songwriter, this didn’t give me much to go by. Perhaps, if I compared the two albums’ most frequently used trigrams?
Aha — now we were getting somewhere. Where Fearless (right) reinforces my earlier criticism, the trigrams from 1989 — namely, the song “Shake it off” -focus more intensely on Swift herself. As she explained to Rolling Stone in 2014: “When you live your life under that kind of scrutiny, you can either let it break you, or you can get really good at dodging punches. And when one lands, you know how to deal with it. And I guess the way that I deal with it is to shake it off.”
Ultimately, my textual investigation should have supplemented a broader investigation which also examined songs Swift wrote vs. co-produced and weighted the popularity of the songs. From the data I did collect, it seems Camilla Pagalia and I should maybe give Swift another chance: the pop star is shifting tone, however incrementally, from the lovestruck ballads of albums past.
Whales, whales, whales
A few weeks ago a friend of mine shared this image that a friend of hers had originally posted to Facebook. The image was not linked to an article and did not cite a source (I have since found that it came from The Sun. The image sent me down a rabbit hole learning about whale beachings (there have been two large ones since the start of the year one of a pod of sperm whale in the North Sea and the other of a pod of pilot whales of the coast of India.
Some articles posed theories about how and why these animals were beaching but most said there were no conclusive reasons cited yet. It seems that it conducting complete narcopsies for whales is timely and expensive. The reports for 21 pilot whales beached in Scotland in 2013 were just released in the end of 2015. That report supported the most likely theory that I had read among the different articles: that the whales had ingested so much mercury over their life times that it had damaged their ability to navigate the waters and resulted in their fatal disorientation. Most papers reported that sperm whales beached in the North Sea had gotten lost in shallow waters looking for a giant squid and noted that this is often thought to be the reason that whales beach: they get lost in shallow waters and then can not get out or can not find food and die before reaching the shore.
But I wondered why the whales were getting lost and if they were getting lost more often then before. Wikipedia offered a listing of all the reported beachings of sperm whales since the mid 1700’s but when i graphed this data it seemed erratic. Then I decided to graph all the reported mass beachings of pilot whales and the steady increase was much more evident. I dropped the sperm whales from the exercise and decided to focus my data on the pilot whales.
The studies are still inconclusive that increased mercury levels cause neurological damage and disorientation specifically in whales but this damage has been proven conclusively in other high order mammals and one article in National Geographic cited the study of the pilot whales and referenced the possible link between the toxins and the beachings.
As further context I visited the New Bedford Whaling Museum as part of my research and had a nice time talking for a few hours to a docent there. The museum seemed to target elementary school programs and I think a bit of that aesthetic seemed into my video!
The map that appears in the video is originally from this site.
What Poverty Steals
I’m fascinated by the tangle of life expectancy, wealth and poverty, income inequality and social mobility.
My data viz was prompted by new research detailing poor people’s shorter life expectancies and The Atlantic article about the 47 percent of cash-strapped Americans who said they couldn’t come up with $400 for an unexpected expense.
Economic stability is about income, but it’s also about assets and wealth. It’s about having a cushion to shield you from the inevitable unexpected expense of car repairs or a medical emergency.
Poor people don’t have that margin of error, which is one of the reasons that economic mobility is so low. Kids who are born poor in Shelby County (the county that holds Memphis) die poor. Only 2.6 percent of children raised in the bottom quintile of household income in the Memphis area rise to the top quintile by adulthood. According to a New York Times interactive, “Shelby County is very bad for income mobility for children in poor families. It is better than only about 9 percent of counties.”
Since Shelby County is majority black and a disproportionate share of the poor people are black, I wanted to focus specifically on black people. (Hispanics also have an insanely high poverty rate, but there’s relatively few of them in Shelby County/Memphis and most are recent immigrants.)
Here’s what I wanted to determine for people in the county where I live, Shelby County:
If poor people had the same life expectancy of rich people, how much more could they expect to earn over those additional working years? If you add up all those dollars, how many millions of dollars are poor black people in Shelby County forfeiting simply because they’re poor, black and live where they do?
If I could answer this question, I wanted to show the data similar to how Periscopic animated the years lost to gun deaths. http://guns.periscopic.com/?year=2013
Spoiler alert: I don’t have the data to answer the question I was trying to answer. Especially not on the $$ end.
Nevertheless, I did a short video, 1:06. And I figured out how to add music.
I was going to build a little bar chart showing the difference in life expectancy between poor people and rich people in Shelby County – or one comparing the income disparity in life expectancy by the biggest counties in Tennessee, but there wasn’t a whole lot of difference. And I can do a bar chart, so I was trying to figure out what I didn’t know how to do.
*** I’m pretty sure my math is all wrong, because there’s far more than 26,000 black adults in Shelby County who are poor (as defined by living in the bottom quartile of household income). Would love to think through how to answer my question with someone who knows.
How to win a Nobel
Last fall, I scraped and cleaned data for the more than 21,000 nominations submitted for Nobel Prizes between 1901 and 1966 — the only years for which data were publicly available. For each nomination, the database contains the names of both the nominator and the nominee, along with such information as their gender, hometown, birth year, death year, and profession.
Some surprising factoids began to jump out at me as I looked over the data. I thought I’d tell the story of one of them for this assignment, .
To see the story, download the zip from www.github.com/aguynamedashley/partnews and open index.html in your browser.
Profile (belated) and data story: Sands Fish
I interviewed Sands Fish for our class profiles assignment months ago and decided to try to profile him through the medium in which he is an expert: data visualization. However, I ran into a road block that I wasn’t able to resolve until our data visualization class. So I’m combining two assignments in one and finally presenting my results.
After Sands and I talked, I transcribed 25 minutes of our interview, including even the “um”s and “yeah”s. Then I analyzed the text from several different perspectives, trying to echo Sands’ work with MediaCloud, which crunches massive amounts of data to discover the relationships between words and the people who use them. In our case, I wanted to get a visual representation of the themes and rhythm of our interview.
First, I analyzed the language we each used. Here are the words I used most often:
And the ones Sands used most often:
There wasn’t a lot of overlap.
Then I counted the number of words in each uninterrupted chunk of speech and made a spreadsheet recording each of those chunks under our respective names, with the minute timestamp interspersed. For example, here is the first five minutes:
Here is a streamgraph that shows our individual share of the conversation, and the overall give and take. I used total words per person per minute to produce this graph on raw.densitydesign.org:
Then I took a more granular look at the first 10 minutes of conversation, using cumulative word count instead of minutes as the x-axis value. That gave me a better sense of the frequency of volleys between us, and the duration of each uninterrupted chunk of speech:
Here are a few takeaways I gleaned about my interview style by representing the interview visually:
- I affirm understanding in lazy ways (yeah, OK, mhmm), and I interrupt a lot.
- It would be better would be to remain silent until the end of my interviewee’s explanation, and then affirm my understanding in a summary that uses key words and phrases that he or she has shared.
- Overall the share of conversation is roughly appropriate for interviewer and interviewee, though the spike at 22 represents a story I shared that probably didn’t add much to the interview.
The American Community Survey – 3 in 1: explainer, engagement, data story
I have thought about creating a census fan page many times. Looking at data all day makes one appreciate the history, scale, and effort of this massive public endeavor. Not only does the census provide official guidance to the formulation of public funding and policy, it has over the years also ritualistically structured our understanding of our environment. Since 1790, the census evolved not just to adapt to the massive increase in population(from under 4 million to 318 million today) and migration(from 5.1% urban to 81% in 2000), but its format has also changed to reflect our attitudes. In this 3 part(hopefully) assignment/makeup assignments, I focused on explaining and visualizing the American Community Survey(ACS), a newer data offering of the census that is a yearly long form survey for a 1% sample of the population.
Last summer, while interning at a newsroom, I built a twitter bot based on the ACS inspired by how nuanced and evocative the original collected format of the dataset is. Each tweet is a person’s data reconstituted into a mini bio. In the year since, people have retweeted when an entry is absurd or sad, but most often when an entry reminded them of themselves or someone they know. It quickly became clear that narratives are more digestible than data plotted on a map. However, I was at a loss on how to further this line of inquiry to include more data in bigger narratives.
Part of my research is to experiment with ways of making public data accessible so that individuals can make small incremental changes to improve their own environment. Many of these small daily decisions are driven by public data, but making the underlying data public is not always enough. While still plotting data on maps regularly, I started to think about narratives. Can algorithmically constructed narratives and narrative visualizations stand alone as long-form creative nonfiction?
There are so many wonderful public data projects that go the extra step out there. Socialexplorer does a great job of aggregating the data, so does actually ancestry.com. Projects from timeLab show many examples of how census data has been used for a variety of purposes, even entertainment. And just last week, the macroconnections group unveiled a beautiful and massive effort to expose public datasets with datausa.io that takes data all the way into a story presentation.
Constraints are blessings…
It’s fortunate that I work in such a time and environment but also very intimidating. What can I contribute to an already rich body of work where each endeavor normally requires many hours and even months of teamwork, not to mention the variety of skills involved? More selfishly, what can visual artists add to the conversation that is beyond simply dressing up the results? This series of 3 assignments is a start.
1. Explainer – the evolution of the census
Instead of focusing on how the population has changed, here is a visualization of how census questions have changed to reflect the attitudes and needs of the times. Unfortunately this was unfinished and only goes from 1790 to 1840 right now.
view closeups here – 1790_1840
2. Engagement – how special are you?
I have been procrastinating by spending a lot of time on guessing the correlation. I think that buzzfeed-type quizzes are one of the best data collection tools. Of course there is also this incredible NYT series. People who commented on the census bot often directly address tweets that describe themselves. This is an experiment to get people to learn something about the data by allowing them to place themselves in it.
This is also still very much in progress: http://jjjiia.github.io/censusquiz/
3. Data Story
To be continued …
Parity Pool
Storytelling with data requires patience, reliable sources, and creativity. I was excited to browse the aggregated data sets on the newly launched DataUSA.io website. I soon found myself lost in statistics about occupations, income distribution, and wage gaps in the United States. Ultimately, I decided to explore educational data provided around Computer Science degrees programs. I wasn’t exactly sure what I would fine, but I new I wanted to look at issues of diversity within the technology sector. Visit http://partnews16-722286.silk.co/ to see what I discovered.
Visualizing Comments by Gender at the NYT
I recently read Emma Pierson’s study about commenters and gender at the NYT. I thought it was a great piece with compelling data, some of which I tried to pull out in the following infographic.
A few challenges: the program I used didn’t allow me a lot of flexibility in terms of editing the charts, so I had to be creative about which points I chose to pull out of her findings. This visualization also uses word clouds, which some folks find terribly unsophisticated, but I really liked the visual comparison of the types of words that men and women use in comments on the same articles side by side.
Without further ado, here’s the visualization…(unfortunately, I had to paste in a screenshot because the original png file wouldn’t copy into this, so the quality on this version is a little lower than I would have hoped)
