My Last Relationship Produced More Words in WhatsApp Than 'Moby Dick'
After a breakup, all we are left with is content. I ran mine through sentiment analysis.
Image: Awe Inspiring Images/Shutterstock
After a breakup my first instinct is to unfollow and unfriend. To the revenge-happy mind of the newly single it's a victimless crime. There is a simple pleasure in channelling all your frustration into a click, one which will statistically affect someone in a way they'll likely pretend they haven't noticed. You tell yourself they'll never see, then cross your fingers and hope that they do.
I've written previously about how technology both facilitates and warps relationships, turning us into Tamagotchi pets which can be switched off at will. You turn into the chatbot of yourself, a soothing girlfriend figure who knows by rote what to tell people to make them happy.
Then after a relationship built largely on messaging ends, what we are left with is words. Less romantically, we are left with "content." Least romantic of all, we are left with data.
My own recently-ended relationship produced 304,101 words of WhatsApp conversation over three and a half months. That's six and a half Great Gatsbys.
To conduct a relationship through technology, even privately, is to welcome strangers into it: You may not tick "engaged" or "married" or "it's complicated," but somewhere on a server they are linked to your profile. "Big data knows you're pregnant," one headline last year read, alluding to how statisticians can create highly accurate profiles of our offline selves determined by our online behaviour. Your romantic life is part of the blueprint of your online being, another determining factor in how you'll vote, how you'll spend your money, and how prolifically you'll use a website's services.
Which is a lot, it turns out. My own recently-ended relationship produced 20,425 lines and 304,101 words of WhatsApp conversation over three and a half months, averaging 194 lines of text per day. To get an idea of volume, I measured it against the word count of famous novels: The Great Gatsby is 47,094 words long, and 144 pages. My relationship produced six and a half Great Gatsbys. Moby Dick, meanwhile, has 206,052 words, depending on the edition. So it produced 1.48 Moby Dicks, though I can estimate that the content was about 0.2 percent as eloquent.
Of course this is all just mindless, data-facilitated self-indulgence. But it's tempting, when companies make the option to reclaim your data available. In 2012 Twitter launched its archive service, allowing users to download their history of tweets. Similarly, WhatsApp has the option to download old chat logs, Facebook lets you download your profile data, and this week Google announced the option to download your entire search history.
In every case, these options leave us with unwieldy, unlovely text files which hold a morbid appeal. What does one do with all that data? Is it really of benefit to look back? In the past, writers and public figures had their correspondence archived and published after their deaths; Hillary Clinton famously ran into trouble, lately, for deleting emails sent while she was Secretary of State. But are our own mundane messages worth keeping, even just for ourselves?
Words are cheap these days: 19 percent of Facebook's users send private messages daily, and the average WhatsApp user sends over 1,200 messages per month. Every day we drown in words, pressured to be articulate as much as we are to look good in selfies.
Part of me believed that seeing three and a half months of my life distilled into data would help me put that time period to rest. Another part of me wanted to go back and see what I could learn from it.
The words "sex," "work," "tough" and "shit" appear prominently. The automated gods are trying to tell me something.
"Sentiment analysis" occurs at the intersection of automated calculation and human emotion: It refers to the use of natural language processing and computational linguistics to extract information from data. At its simplest it can process words—tweets, messages, news coverage—and tell you if they swing positive or negative. At its most complex it delves into human nature, identifying hostility and cynicism, subjectivity or objectivity. It played a central role in Facebook's notorious "emotional manipulation" experiments of last year. It reads the room for marketers and politicians.
And now I'm going to treat the cloud as my personal magic 8 ball, using free tools to find out more about myself and my failed relationship.
The simplest tool to begin with is a word cloud: You feed it a chunk of text and it creates a data visualisation highlighting your most frequently-used words. Given my WhatsApp chats with my ex-boyfriend, the words "sex," "work," "tough" and "shit" appear prominently in various clouds. The automated gods are trying to tell me something.
Some sentiment analysis tools resemble a Love-ometer at a fairground—outdated and clunky Web 1.0 sites, the internet equivalent of "vintage." The "Free Sentiment Analyzer," hosted on the personal website of a systems analyst from California called Daniel Soper, gives my relationship a score of 26.8 ("somewhat positive/enthusiastic") which I read as grim. It takes a homemade robot to inform me that my life lacks passion.
The most sophisticated analysis I came across was provided by Watson Personality Insights, part of an IBM-backed developer cloud community where cognitive computing tools are publicly shared. Personality Insights can be used in customer acquisition, marketing strategies, and even the writing of personal CVs. It relies on a series of personality models, gauging "needs," "values," and the "big five" characteristics: agreeableness, conscientiousness, extraversion, emotional range, and openness. You paste in your text, and it gets to know you, or your customers, almost instantly.
I feed the robot oracle my three months' worth of chats and it gives me a reading: 48 percent openness, 63 percent conscientiousness, zero percent extraversion, 82 percent agreeableness, 100 percent cautiousness, and zero percent "excitement seeking." Agreeable but unexciting, even by robot standards. Apparently we were doomed from the start.
It gets worse—scroll down and you receive a densely-detailed pie chart colour-coded in red and blue. Apparently we had one percent "hedonism" and six percent "cheerfulness," though reassuringly there was 71 percent "closeness."
Craving reassurance, I switch to personal analysis of my own name through a site called opinioncrawl.com, a broader sentiment analysis tool where you search names and single words. I type in my name and the site throws up several pictures, one of me and two of people I know, and another grim pie chart which tells me I generate 54 percent negative attention (fuck you, readers!). My "key concepts" in this dismal internet portrait include "advertising," "budget," "breakfast," "misogyny problem," "illness" and "Sudoku," which I do not recall ever having written about.
I want to throw my laptop out the window, curl up in bed and cry. Oh data, what have you done to me? Why on earth would anyone want to see this?
And yet sites make it easier to see our unflattering data reflections everyday. And personal archiving is apparently in fashion—Facebook's "On This Day" function, introduced last month, only encourages digital navel-gazing, while companies like Memeoirs and Chatbooks allow you to convert old chat logs and pictures into printed books and keep them forever.
That said, I wouldn't advise it. Think of all that time you could spend reading Moby Dick. The next time I go near sentiment analysis tools, I think it will be in the middle of a relationship, rather than at the end of one. If I'd known what I was writing in my messages was only "one percent excitement-seeking" and "six percent cheerful," maybe I would have done things very differently.