Life in your early to mid-20’s likes to play tricks on you in the fact that as soon as you feel comfortable in a new setting it ships you out to somewhere else. I’ve spent my last 8 weeks in New York, Texas, Florida, and finally after a long 5 days of being back in the city (the longest I’ve been in one place since June!), I get to go home. And by home I mean Pennsylvania, and then about two weeks later I get to go back to real home-away-from-home, a.k.a. school. So really, to quote a famous 21st century poet, life is literally saying, “We Can’t Stop”.
But otherwise, life’s been pretty fun! In the past 8 weeks I’ve met some pretty amazing people, had the opportunity to work on a pretty sweet healthcare project, and also spent my weekends in one of the most amazing cities in America, surrounded by old friends, new friends, family, jazz cafes, and oddly themed bars (ask my friend Gus about a small place called Moscow 57 on Delancey St…don’t ask him how or why we ended up there).
Now when you’re being tossed around the country, lack of sleep catches up with you. This is not something uncommon as a college student, but until recently I’ve never been successful at sleeping on the go…it’s a skill you have to learn. But on plane flights at 6:24 in the morning (my Monday weekly), it doesn’t become too difficult to throw your head back and catch some zzz’s. Now, roll forwards to New York City Saturday nights; nights spent on subway cars, corner bars surrounded by crowds of people in dark rooms and basements with smoke-filled air. Ask my friend Maddie about these nights – before you realize it they’ve gotten away from you, and the clock strikes 4AM and you rush home waiting for a train in a humid subway stop, half asleep on 14th and 8th with Thomas Nast statues lurking in the distance.
Many of these nights I would go visit my brother, who used to live in a typical hole-in-the-wall Williamsburg apartment (though it did have character). The type of apartment you would expect 3 young professionals in New York to live in – and in any other city would fit one. To do this, I would take the A train down to my favorite station above, and then use the L to reach Brooklyn. It was on one of these late nights, that I woke up from a normal daze on the L-Train to look above me and see a very colorful add neatly pasted above the dark New York subway windows. It was from a small company called Casper sleep and here are a few examples:
Other than the irony of waking up to a mattress add on a 4AM subway train (and boy did my mattress sound wonderful at the time), I really enjoyed the playful nature of these pieces. I would look forward to my time on the L-Train, where I could study the cute cartoons and feel a little comic-relief to whatever was on my mind at that time of the day, and I began to wonder what attracted me to the tiny adds. Was it the color-scheme? Bright, yet faded with a plain blue line-distinction and soft shadowing effects. Or, maybe it was the pictures. What person doesn’t like cute animals personified by eating pizza, or sleeping cozily next to each other. Or maybe it was the sentence below, using words that usually are associated with happy subjects: “mattress”, “perfect”, and words that represent friends and comfort like “locals” or “frenemies” (which is a cute antithesis in itself).
The idea of taking subjective qualitative concepts, such as an emotion described with the words of an advertisement, and trying to generate objective subject-matter around it is a unique topic in the field of machine learning. Now “machine learning” might sound like some big, dangerous topic, but in reality all machine learning involves is training a machine (like a computer) to make some recognition that normally a human would make. It allows computers to recognize faces in pictures, Siri to talk to you on your iPhone, and political scientists to interpret election data to make a prediction for the future.
Ask Siri what 0/0 is and you will understand why this is here.
Now comes the fun (heh heh evil engineering laugh…ok back the post). I’m going to show you one application of machine learning, and how it might be able to explain my fascination with Casper sleep adds. There’s one topic in machine learning called Natural Language Processing, or simply NLP. Essentially, NLP is a topic that is trying allow computers to interact with human language – a subject that can surprisingly be difficult to derive information from, because the meaning of a sentence can be a really subjective thing. Think about it: when someone speaks to you, the meaning you derive from a sentence is largely dependent on the situation surrounding it, the speaker, and then the word choice used. It’s not easy to program this stuff for a computer to recognize.
So how is this useful for marketing? Well, there’s a topic in NLP called sentiment analysis. Sentiment analysis deals with trying to pull subjective information such as text, and group them into categories, maybe emotional categories. Let’s say we have a group of text from a Casper Ad, like the following:
“Perfect” “Mattress” “Locals”
As you can see, I’ve stripped this text from the above add of all articles and prepositions, to really pull parts of speech that deal with larger subject matter. The idea for a lot of NLP is to try and build dictionaries of relevant words that fit these categories, and then try to label words with these categories. For example, let’s say I have two categories, a category for Positive words and a category for Negative words, and I have dictionaries of words for these categories. I can write software that tells a computer to scan in text from an advertisement, and label all relevant words that fit these categories. For instance if I were to feed in the three words above, they might be identified to fit positive and negative categories as follows:
Positive: mattress, locals
This means that the words “mattress” and “locals” were found in the positive dictionary, and “perfect” was found in the negative dictionary. Now, what happens if words are not in the dictionary? Well, then there’s a little manual work involved to identify words and place them in the categories yourself. But, as with most machine learning, the idea is that the more advertisements one shows a computer, the more words will be placed into a dictionary, and in the long-run the majority of words will be labeled, thus less manual work will be involved.
So, great, computers can build dictionaries…but just word identification itself is not really helpful. So how is this useful for marketing? Well, NLP can be used as a great feedback tool for advertisements. For example, let’s say that there is a specific tone in an add that a marketing firm is trying to get across. How can they find out if this concept is true? One method might be to look at Facebook commentary to that add, and see if, for instance, the comments in the add are identified by a computer as falling in the same tonal category for the add itself (and a computer can go through thousands of comments much faster than we can). Or, maybe we pair adds with the amount of Twitter favorites they create. One could see whether adds that fall into a certain tonal category tend to generate more favorites…and therefore one might state that adds that create certain tones are more likely to be attractive. So yeah, using data analysis and with language processing can be pretty cool stuff.
Or, it can put you asleep (figuratively…if this is boring enough, literally…if you’re me on a subway car). But regardless, it gives more power to how words can send overall impact and responses can be measured. This is powerful in communication as a whole, something we all deal with on a daily basis. But, back to sleep, which is something I’m about to engage in quite comfortable on my flight home in about an hour. And with that…hope everyone has had a well rested summer, because sometimes there’s nothing more beautiful than a pillow, mattress, and catching some zzz’s at home.
Until next time…peace out NYC