Skip to Content
Banner image: A woman successfully wins a game of Wordle. (Credit: Glenn Asakawa)
Created by Josh Wardle, the game Wordle went public in October 2021 and has since skyrocketed in popularity and inspired many offshoots. It’s surprisingly simple: Each day, players have six attempts to guess a randomly generated five-letter word.
This free online word puzzle became so well-known over the winter that it was recently purchased by The New York Times Company, setting off a flurry of debate. Would the game become more difficult, or would you have to pay to play?
Alexis Palmer, assistant professor of linguistics at CU Boulder, is a self-described word game aficionado. She is also a computational linguist, a subfield of artificial intelligence that sits at the intersection of linguistics and computer science. CU Boulder Today spoke with Palmer to learn more about the Wordle phenomenon.
Alexis Palmer, assistant professor of linguistics at CU Boulder
Five-letter words are interesting. They’re usually just one syllable in English, and usually they start with consonants and their vowels land somewhere in the middle. Another common structure for five-letter words is alternating consonants and vowels. These are constraints that we know innately and we use that knowledge whether we realize it or not. It’s interesting as a window into linguistics, because as you’re solving this, you’re like, well, I know what letters can go there. How do I know that?
Wordle appeals to our intuitive knowledge of what words can look like in English. Not only what words are in our vocabulary, but what words are possible. There’s a huge amount of knowledge about valid word structures and patterns of sound that we have on a subconscious level as speakers of our native language. Even if we couldn’t write down the rules, we know you can’t start a word with the sequence “ng” in English, that’s just never going to happen. This is a lot of what linguistics is: trying to codify those rules and figure out what our knowledge as a native speaker is.
There are a lot of strategies that are just based on the frequency of letters in English. This also shows up in Wheel of Fortune where the game starts contestants off with a certain set of letters—R, S, T, L, N and E—because those are the most frequently occurring letters in English. A lot of the words that have been proposed as good starting words are some combination of those.
With Wordle, I actually first started with a strategy of trying to figure out what my vowels are. I use a starting word, “LOUIE.” According to something I read this morning, that’s a bad strategy, but I also haven’t lost yet, either. Of course, from a computational perspective, you could also just write an algorithm to solve Wordle puzzles, but where’s the fun in that?
It’s pretty simple on the back-end. There are two different word lists that are involved. One is the list of the potential words of the day. That list is a little over 2,000 words. They’re randomized and they move through a pre-specified order. The second word list is the list of valid guesses, and that’s over 10,000 different words. Even Josh Wardle, the guy who created this version of the game, doesn’t know what the next word is going to be, so that he can play it.
The New York Times made a couple of changes, but only removing words, not adding any words. They removed some words for appropriateness, and then they also removed a few words they thought might be obscure or reflect British instead of American spelling. So, no. They did not make it harder.
There’s a Canadian linguist and software developer named Aidan Pine, who put up an open source version of the Wordle code that lets you create a new Wordle for any language. He’s got the repository of code on GitHub, and also a blog post that explains in great detail how to do this. It’s written for people who might not have any software development background. Anyone who’s a speaker of any language can build their own Wordle.
Part of my research is looking at how we can use computational methods to support documentation of endangered languages, to support the work of revitalization of these languages. And these kinds of games can be a really great tool for engagement with the language, both for speakers and for people who want to learn the language. Pine has built Wordles for two different First Nations languages spoken in British Columbia. Some of the speakers of that language have talked about how it’s really helped them—one woman says she’s made a commitment to spend at least 15 minutes a day working on learning her ancestral language.
Learn more
Learn more about computational linguistics and its degree programs at CU Boulder.
I wouldn’t be surprised, honestly, to see Wordle have the kind of longevity that crossword puzzles do. It’s such a satisfying activity, with the additional fact that it relies on our own internal knowledge. It doesn’t have anything to do with our level of education or what books we’ve read or things like that. It’s really all about our internal knowledge of language and things that we inherently know, whether we realize that we know them or not.
Computational linguistics/natural language processing is an inherently interdisciplinary field, and it informs a hugely important set of technologies these days. When I started in this field maybe 20 years ago, it felt like it was one of the really important frontiers, and now it’s become part of our daily life. Language technology, for example, ranges from things like spell checking and predictive text and games like Wordle, to things like Siri, Alexa and Google Translate.
All of the big tech companies like Google, Amazon and Facebook are hiring people who know these kinds of technologies. But they’re also hiring linguists, because linguists have really great analytical skills to analyze what language is doing, figure out what the patterns are, and figure out how we could improve these sorts of systems.
Subscribe to CUBT
Sign up for Alerts
Administrative eMemos
Buff Bulletin Board
Events Calendar
CU Boulder Today is created by Strategic Relations and Communications.
Campus images reflect public health requirements at the time of the photograph or were taken prior to the COVID-19 pandemic.
Faculty-Staff Email Archive Student Email Archive Graduate Student Email Archive New Buffs Email Archive Senior Class Student Email Archive Community Email Archive COVID-19 Digest Archive
University of Colorado Boulder
© Regents of the University of Colorado
Privacy • Legal & Trademarks • Campus Map
