Not So Loud

THANK YOU

How do you punctuate around emoticons and emoji?

I’m on Mental Floss talking about an important issue: how to integrate emoticons and emoji into our system of punctuation: 

🆒🆕In October 2011, Apple added the emoji keyboard to iOS as an international keyboard. Since then, digital language has evolved such that nearly half of comments and captions on Instagram contain emoji characters. And earlier this week, Instagram also added support for emoji characters in hashtags, which allows people to tag and search content with their favorite emoji #🎉.

In Part 1 of this blog post series, we will take a deep dive into emoji usage on Instagram. By applying machine learning and natural language processing techniques, we’ll discover the hidden semantics of emoji.

Emoji on Instagram: Up and to the Right

It is a rare privilege to observe the rise of a new language. Instagram has always supported emoji, but they did not see wide adoption until the introduction of the emoji keyboard on iOS (October 2011) and on most Android platforms (July 2013). The graph below shows the percentage of text (comments and captions) containing emoji characters graphed over time 📈.

In the month following the introduction of the iOS emoji keyboard, 10% of text on Instagram contained emoji. The trend continued until the release of Instagram for Android in April of 2012, when many new users did not have emoji support. Afterwards, there was a clear upward trend which accelerated after Android received native support for emoji in July 2013.

Usage continued to grow and in March of this year, nearly half of text contained emoji 😱. In the future, will all text contain emoji? To help answer that question, we divided emoji usage by country and observed the differences between user cohorts.

The graph below shows that users from Finland are using emoji characters in over 60% of text! In contrast, the lower bound is in Tanzania with only 10% of text containing emoji. If the overall trend continues, we might be looking at a future where the majority of text on Instagramcontains emoji.

Natural Language Processing

Learning an Emoji Representation

We’re often asked about the meaning of emoji such as 🙇. Intuitively, substitutable words have similar meanings. For example, we might say that “dog” and “cat” are similar words because they can both be used in sentences like “The pet store sells _ food.” In the field of natural language processing, this intuition is called the distributional hypothesis 🎓. It can be applied to emoji by treating them as if they are normal words.

More formally, we can place (or embed) emoji and hashtags together with words into a common metric space where there are well-defined distances between elements. The representation of the words are chosen so that similar words have a small distance. In the scatter chart below, we embedded words, emoji, and hashtags into a 100-dimensional space of floating point numbers using 50 million English Instagram comments and captions from 2015.

We learn the floating point numbers using the Gensim library, which re-implements a tool called word2vec. In skip-gram mode, word2vec reads through text and predicts the context around a given word or emoji. If the algorithm predicts the context incorrectly, then it adjusts its internals to make a better guess in the next round. As part of that unsupervised training process, word2vec learns our 100-dimensional representation for words and emoji.

Emoji Translations

Having learned a good representation for emoji, we can begin to ask questions about similarity. Namely, for a given emoji, what English words are semantically similar? For each emoji, we compute the “angle” (equivalently the cosine similarity) between it and other words. Words with a small angle are said to be similar and provide a natural, English-language translation for that emoji.

Using our algorithm, we find that many of our popular emoji have meanings in-line with early internet slang:

😂 (ranked 1st in emoji usage): lolol, lmao, lololol, lolz, lmfao, lmaoo, lolololol, lol, ahahah, ahahha, loll, ahaha, ahah, lmfaoo, ahha, lmaooo, lolll, lollll, ahahaha, ahhaha, lml, lmfaooo

😍 (ranked 2nd in emoji usage): beautifull, gawgeous, gorgeous, perfff, georgous, gorgous, hottt, goregous, cuteeee, beautifullll, georgeous, baeeeee, hotttt, babeee, sexyyyy, perffff, hawttt

❤ (ranked 3rd in emoji usage): xoxoxox, xoxoxo, xoxo, xoxoxoxo, xoxoxoxoxo, xoxoxoxox, xxoo, oxox, babycakes, muahhhh, mwahh, babe, boobear, loveyou, bunches, muahhh, muahh, xoxox, muahhhhh

👍(ranked 9th in emoji usage): #keepitup, #fingerscrossed, aswell, haha, #impressed, #yourock, lol, #greatjob, bud, #goodjob, awesome, good, #muchlove, #proudofyou, job, #goodluck

😭(ranked 11th in emoji usage): ughh, ughhh, ughhhh, ugh, uggh, ugghh, ughhhhh, ughhhhhh, ugggh, lolol, wahhhh, rn, oml, uhg, agh, xc, omgg, omfg, omf, lololol, whyyy, loll, wahhhhh, tooo, kms

Some of the more distinctive emoji had particularly distinctive meanings:

🙌: #waitonit, #justwaitonit, #wonthedoit, #nuffsaid, #yeslawd, #youtherealmvp, #stayblessed, #thatisall, thou, #enoughsaid, leggo, #onlythebeginning

👯: #sistasista, #sistersforlife, #sistersister, #bestiesforlife, yearsoffriendship, #sisterfromanothermister, #morelikesisters, #bffl, #bestiesfortheresties, #bestfriendsforever

💃: #birthdaybehavior, #bdaybehavior, #tu, #ladiesnight, #turnuptime, #dontmissit, #bdaycelebration, #piscesseason, #bethere, turnup, #grownandsexy

🎅: merry, christmas, #merrychristmas, #christmas2014, #christmaseve, #christmastime, xmas, eve, #santa, claus, #happyholidays, #xmas, clause, reindeer, pesach

Naturally, people have strong associations with the flag emoji:

🇺🇸 : merica, #godblessamerica, ‘merica, #murica, #merica, #hooah, #america, #specialforces, #supportourtroops, #goarmy, #redwhiteandblue

🇫🇷 : paris, france, #eiffeltower, #paris, #france, louvre, italy, #montreal

🇯🇵 : #japan, #osaka, #kyoto, #japanese, japan, taipei, osaka, beijing, taiwan, tokyo, #日本

And in answer to our question, we can find that the 🙇 emoji is associated with: #goodmorningtho,#yadigg,lbvs,#gn,#inmyfeelings,#latenightthoughts,#deletinglater. Personally, I like laughing but very serious (lbvs).

Changing the vocabulary

It seems that the most popular emoji have similar semantics to words like “lol/hehe” (😂), “xoxo” (❤️) and “omg” (😱). Are these emoji also replacing the usage of the words?

Precisely, we examine the usage of language in Instagram comments and captions by measuring the percentage of text containing emoji or internet slang. To control for natural changes in Instagram demographics, we examined four cohorts past the launch of Instagram for Android: those joining Instagram in the first week of July 2012, January 2013, July 2013, and January 2014. Each cohort contains millions of Instagram users. We defined internet slang as words matching variants of “xoxo”, “omg”, “muah”, “babe”, “bae”, “lol”, “haha,” and “hehe” with the following regular expression:

(?:\b|#)((?:xo)+|omg+|muah+|babe+|bae+|lol+|(?:ha|he)+h?)(\b|\.|!|\?)

As shown in the chart below, all groups exhibit a similar pattern in the rise of emoji (with an upper bound around 45%) and a decline of internet slang (with a lower bound of around 5%). Correlation coefficients within the respective cohorts are all below -0.93, indicating a strongly negative correlation.

The vocabulary of Instagram is shifting similarly across many different cohorts with a decline in internet slang corresponding to rise in the usage of emoji.

The Hearts of Instagram

Having our vectorized representation opens up a wealth of semantic analysis. One of the purported advantages of word2vec representations are that they allow for algebraic operations in semantic space. For example, it can be particularly hard to distinguish the heart emoji 💙💚💛💜💖💗💌. We can isolate some of the effects by subtracting off the representation of ❤️ and finding similar concepts roughly corresponding to color. For example:

💙 - ❤️ ~= #goblue, #letsgoduke, #bleedblue, #ibleedblue, #worldautismawarenessday, #goduke, #beatduke, #autismspeaks, #autismawarenessday, #gobroncos, duke

💚 - ❤️ ~= #gogreen, loyals, #herballife, #happysaintpatricksday, 🍏, #stpats, 🍀, #jointhemovement, green, #hairskinnails, #happystpatricksday

💛 - ❤️ ~= 🌱 ,🍊 ,#springhassprung ,🔆 ,#springiscoming ,#springishere, #aprilshowers, #thinkspring, #hellospring, 🌻, #wildflower, #happyearthday

💜- ❤️~= ✨, 🌀, 🔮, 🌟, 💄, 🎀, faldc, 💎, brassy, topaz, peachy ,purple, #thinkpink,☁, sparkle, 🌿, shimmer, sparkles, kaleidoscope, periwinkle, 🍄, greenish

💖 -❤️ ~= gorl, 💮, cwd, s4s, aynmalik, spvm, ulee, 💧, 🈹, yulema, sfs, bvby, ɑnd, indirect, priv

💗 -❤️ ~= ulitzer, 🎀, peachy, february’s, tulle, mackz, kendall’s, curvy, faldc, #dancewear, strapless, 👗, ◽, floral

💌 - ❤️ ~= 📫, ℹ, 📬, 📮, ✉, 📩, 💳, 💻, 📦, paypal, 📧, item, ⏬, 📱, inquire, orders, payment, 📄, 📋, 📲, deposit

Naturally, there are some mistakes in this type of algebra. Nonetheless, subtracting off ❤️ often leaves us with events highly associated with a specific color like #goblue, #gogreen, peachy and purple.

A Semantic Map

A hundred dimensions are pretty hard for humans to visualize. To visually inspect the relationships between emoji, we can take our 100-dimensional representation for emoji, reduce it two dimensions, and then plot them on a grid. We do this using an algorithm called t-SNE, which attempts to preserve relationships in a visually meaningful way:

Many clusters emerge: food emoji on the left, opposite the work emoji in the top right. Shoes (bottom right) are associated closely to handbags while bathing suits are closer to the water and marine animals (top left). Alcoholic drinks (bottom left) cluster together with bowling. Towards the center, we see a large clustering of facial expressions bordered by sadness, shock, laughter, happiness and coolness. As we travel downwards, we can see happy, love leading all the way the family and wedding emoji.

One has to be careful not to read *too* much into the representation since it is an attempt to produce a 2D space out of a 100D one. But it’s clear that semantics are being approximated in our representation.

Part-1-ing Thoughts

On Instagram, emoji are becoming a valid and near-universal method of expression in all languages. Emoji usage is shifting the people’s vocabulary on Instagram and becoming an important means of expression: their use is anti-correlated with internet slang like “lol” and “xoxo.” By observing words and emoji together we were able to discern representations of both. These representations can help us better understand their semantics and find distinctive characteristics of similar symbols.

Stay tuned for Part 2 on how we actually implemented emoji hashtags. #😳

Reddy & Stanford (LSA 2014 presentation, in press for American Speech) discuss these forms as newly-formed clippings favored by young urban women, but I don’t see anything from their abstract to distinguish the forms with and without the -s.

Baclawsky (LSA 2012 poster presentation) calls the -s morpheme a register-marker — which seems reasonable enough to me — and posits that “this developing language change derives from other -s morphemes, such as diminutive and adverb-markers. These multiple sources result in the semantics of an in-group register-marker that fills the niche of non-verbal communication in text-based conversation.”

Basically, my assumption is that clipping occurs as a register marker for casual, relaxed speech (“prob”, “whatev”) and -s is a separate one indicating cuteness or smth: “I wants pizza”, “My feets hurt” (and I’m assuming that this is partly what the second presentation was about). Combined, they create “probs”, “whatevs”.

Type 1 abbrevs are formed by clipping a polysyllabic word off at a stressed syllable, such as whatever > whatev, unfortunately > unforch,  or abbreviation* > abbrev.

There are historical antecedents for this sort of formation, of course – prototumblinguist reminded me of the Victorian slang word morbid > morbs, which is a similar if slightly distinct sort of formation, and Starr mentions a few Type 1 instances like delicious > delish from the early 20th century – but she says the current trend started around 2002 and that these abbrevs have been increasing in popularity since 2006. She doesn’t discuss any variable timeline for the three different types, but maybe you’re right that Type 1 forms existed on their own for a while before the others. (On the other hand, she notes that the website whatevs.org dates back to January 2002, which predates your own remembered timeline. But there may of course be some regional variation at play here.)

So yeah. These forms are called abbrevs, linguists *have* been looking at them, but there’s still a lot more to say on the subject! Starr’s got an interesting discussion of how people use them both seriously and ironically (and sometimes in ways that seem deliberately ambiguous between the two), if that’s anyone’s cup of tea.

*her slides actually say abbreviate for this one, but I think abbreviation makes more sense as the root word, since abbrev is (I believe) only used as a noun.

I’m especially intrigued by use #13 – a twitter fav meaning “this is now the end of the conversation,” which I’ve definitely used myself. I’ve also noticed this same use of a like in a facebook comment thread. I’d say it’s less common on tumblr because there’s no threading or separate likes, but I think you can use it sometimes if you’ve got a series of reply or ask posts. 

At any rate, it’s an elegant solution to the fact that we don’t want to bid each other farewell at the end of every short interchange, but it’s still weird to leave someone hanging, wondering if they’ve even seen your last message and if they’re going to reply – a fav and no reply within a few minutes acknowledges that they’ve seen your message and that they could have replied but didn’t, so the conversation is now concluded. 

ok SO. a lot of this comes from various stuff i’ve seen on the linguistics of tumblr, but at the heart of it is that people in my generation (at least in the us; idk abt other countries’ timelines on this front) went thru (or are still going thru) our Formative Social Years in an environment where we’d regularly interact with even our closest friends on text-only platforms (whether texting or gchat or fb messages or w/e), and b/c so much linguistic/social information is actually conveyed by facial expression and tone of voice, we’ve collectively made up all of these textual ways of conveying that in a concise, efficient way

so like, sometimes on this blog i’ll talk about “straight people”, and sometimes i’ll talk about “str8 ppl”, and even tho i would pronounce those the same, the first is much more neutral — it would probably happen in the context like “i’m not sure how i feel about straight people writing stories that center around experiences of homophobia” — than the second, which which is much more frustrated/venting — it would be more likely to crop up in the context of “all i want is to live quietly in my little queer utopia but no str8 ppl have to come along and heteronomativity UGH #over it #whatever #NOT RLLY OVER IT”. or even with more subtle things like end punctuation: “i’m not going” basically just means i’m not currently planning to go to the thing; “i’m not going.” carries much more of a connotation of “i have seriously considered going and have Reasons for staying at home” (and note that capital — “i have Reasons for staying at home” feels different than “i have reasons for staying at home”). (and this isn’t even getting into things like shitposting or advanced memeology, but there are specific textual markers that go with things like that, some of which would be pronounced if you read them aloud, but many of which wouldn’t be)

but, crucially, for these kinds of things to carry meaning, they have to be used consistently: if i use “str8 ppl” and “straight people” interchangeably in all contexts (as i do for something like “the supreme court” vs “scotus”), then there’s no way to develop a distinction in meaning between the two — the only way to do that is to consistently use the different orthographies in different contexts. (to take another example: if something is “great”, then it’s solidly good. if something is “gr8”, it’s more in the land of “i can’t quite believe this is as earnest/tacky/tasteless as it is but i’m weirdly into it anyway?” (sometimes with a side helping of “do i just enjoy this ironically or do i genuinely enjoy it there is no way of knowing please send help”))

the upshot of this is that to be fluent in tumblr (or texting, or fb messenger, or w/e) means to actually be paying a lot of attention to subtle points of grammar and spelling, to know when to use “did u kno” or “ur” or even pull out an old-fashioned tip of the hat to “e733T haxxor 5killz”. most of these are very subtle distinctions, the kind of things you feel intuitively rather than write out explicitly, and so it’s very hard to convey them concisely and accurately to someone who’s not already immersed in the linguistic environment

and let’s be real, people in my parents’ generation aren’t. i mean, sure, many of them have facebook accounts, but these kinds of platforms weren’t around when they were in their “really getting to grips with social interaction” years, and their most important social interactions usually don’t take place exclusively online. for me, all of my closest friends are people i’ve only interacted with online for more than a year now (with a few brief face-to-face visits when various travel arrangements have allowed), so tumblr, facebook, and gchat are absolutely critical to my social life and interpersonal interactions; for my parents, their closest friends are people they see in person at work every day, so social media is a light overlay to their social lives, not the thrumming core

as such, my parents don’t grok these distinctions. to them “what are you doing?” means the same thing as “lol wut r u doing”; “gr8” is just like “great” (and “gr9” takes some parsing … ); dogespeak doesn’t have the same distinctive valence that it does to us. since they don’t know about these distinctions, they don’t feel the need to maintain more “proper” spelling/grammar when texting with a friend — different people have different set points for this, obvs, but in general i feel like “standard (setting aside all the class and racial implications in that term …) spelling and grammar” (with lighter-than-standard punctuation and capitalization) translates to “relatively neutral/pleasant conversational voice”, and then deliberate misspellings, abbreviations, letter substitutions, and grammar deviations are markers used to indicate shifts in mood — i have a vague sense that bitterness tends to collapse down and preserve grammar but weird spelling (“lyk w/e im happy 4 u but pls, i kno u lied 2 get that”) whereas enthusiasm tends to preserve spelling but weird grammar (“what i can’t even no how do air AMAZE”). since people in my parents’ generation don’t realize that doing so unintentionally changes the way their words come across, they feel free to text “poorly” (ie with lots of errors/substitutions, generally mixing various text-flagged vocal tones in ways that are often incoherent) in order to do so more quickly (b/c lbr typing everything out can be a pain (esp on a non-smartphone), and since parents don’t do it as much, they’re not necessarily as fast as our spry young fingers on a familiar interface)

so yeah, that’s what i suspect is going on

no    =    “No,” she said. 

no.    =    "No,” she said sharply.

No    =    “No,” she stated firmly.

No.    =    “No,” she snapped.

NO    =    “No!” she shouted.

noooooo    =    “No,” she moaned.

no~    =    “No,” she said with a drawn-out sing-song.

~no~    =    “No,” she drawled sarcastically.

NOOOOO    =    “No!” she screamed dramatically.

*no* = “No,” she said emphatically. 

*nopes on out of here* = “No,” she said of herself in the third person, with a touch of humorous emphasis.

~*~noooo~*~ = “No,” she moaned in stylized pseudo-desperation.

#no = “No,” she added as a side comment.

“no” = “No,” she scare-quoted.

wtf are you kidding no = “No,” she said flatly. “And I can’t believe I have to say this.”

no no No No NO NO NO NO = "No,” she repeated over and over again, growing louder and more emphatic. 

nooOOOO = “No,” she said, starting out quietly and turning into a scream.

1. How she just tappin random buttons in the beginning

2. How she says “I am confusion”