Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

A 2020 University of Vermont study analyzed how people stretch letters in informal online writing, such as “duuuude” and “hahahaha.” It introduced two ways to describe those spellings—stretch and balance—and proposed uses for language technology. It did not show that training an AI model on the data improved its performance.

What the scientists studied

Tyler J. Gray, Christopher M. Danforth, and Peter Sheridan Dodds examined “stretchable words”: informal spellings that repeat or elongate letters, including “heellllp,” “heyyyyy,” “gooooooaaaalll,” and “hahahaha.” Such spellings can add emphasis, exaggeration, or tone that a conventional spelling may not convey.

Their paper, “Hahahahaha, Duuuuude, Yeeessss!: A two-parameter characterization of stretchable words and the dynamics of mistypings and misspellings,” was published in PLOS ONE on May 27, 2020. The headline’s connection to AI refers to possible applications of the analysis, not to a newly trained commercial AI system.

How many tweets were analyzed?

The authors analyzed roughly 100 billion tweets from a 10% random sample of Twitter’s “gardenhose” stream, spanning September 9, 2008, through December 31, 2016. They limited the analysis to tweets flagged as English or not flagged for any language. This was a large historical sample, not every tweet posted, and it does not describe current social-media writing by itself.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Twitter’s API terms prevented the authors from redistributing the individual tweets. That access restriction matters for anyone trying to inspect or reproduce the analysis from the original messages.

What do “stretch” and “balance” mean?

The study separates two properties that can otherwise blur together:

  • Stretch measures the overall amount of letter repetition or elongation in a word.
  • Balance measures how evenly that elongation is spread across the word’s characters. Repetition distributed across several letters is more balanced than a word in which just one character is extended.

For example, “heyyyyy” concentrates repetition on one character, while “gooooooaaaalll” extends multiple parts of the word. The measurements let researchers describe those patterns without treating all nonstandard spellings as the same kind of variation.

The authors used balance plots and spelling trees to visualize letter patterns and examine how misspellings and mistypings develop. The tools make the structure of a spelling variation easier to study; they do not, on their own, determine a writer’s intent or emotion.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.

Can AI understand “duuuude” or “hahahaha”?

The study offers a way to characterize forms that standard dictionaries often leave out. That can be useful to researchers working on language processing, dictionary augmentation, or search, where a system may need to connect an elongated spelling with a familiar word or recognize that spelling variation carries information.

But measuring a spelling pattern is not the same as proving that an AI understands it. The paper proposes that its measurements and visual tools could support language-technology research; it does not report a benchmark showing that a large language model or other production AI became more accurate after being trained on them.

Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Support on Ko-Fi

What the study contributes—and what it does not

The University of Vermont research team described its work as a way to collect and count stretched words and map them by overall stretch and balance. In a ScienceDaily report, the team said the tools could aid continued linguistic study and areas such as language processing, dictionary augmentation, search engines, and sequence construction.

  • It contributes: a large-scale analysis of letter elongation and a two-measure framework for describing it.
  • It suggests: possible applications for language-processing tools and related research.
  • It does not establish: that an AI system was trained on these measurements, that any system improved, or that the results generalize to every platform, language, or period.

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

Special offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.