The Tartan

Adrien Marenco
Staffwriter

On Wednesday, March 26, the Carnegie Mellon Linguistics Club hosted an event called “Linguistics in Action,” a linguistics research showcase, with assistant teaching professor Christina Bjorndhal. 

The presentations and panel discussion featured former and current Carnegie Mellon students Liza Sulkin, Johann Schechter, Jack Sun, and Estelle Tian.

‘Sounding like a Lesbian’

Sulkin, who graduated with a BA in linguistics from Carnegie Mellon in May 2020, is currently working on a Ph.D. in sociophonetics at Boston University. 

Her dissertation, which is titled “Sounding like a Lesbian,” is focused on identifying the “lesbian speech stereotype” that has been conflictingly reported on in past linguistics research. This stereotype, Sulkin explained, is characterized by a low pitch and relatively monotonous voice, though different studies found different results as to the severity of this difference between lesbians and straight women. 

To analyze this phenomenon, Sulkin created an interview-based analysis of different assigned female at birth (AFAB) subjects, coding the participants according to their self-reported presentation, with 1 being “hyper-feminine” and 5 “hyper-masculine,” mirroring the identities “femme” and “butch,” respectively. 

These interviews sampled participants’ speech and included participants who were fluent in English. She found that, as a speaker presents more masculinely, they tend to have a lower pitch. This relationship was mirrored by the speaker’s degree of monotony, albeit with a slightly less strong correlation. 

In conclusion, Sulkin found that her study supported the existence of a lesbian speech stereotype, but one that is dependent on appearance and behavior-based identities in the community. 

Discord emotes and syntax

Next up was Johann Schechter’s analysis of the “Syntactic behavior of discord emotes,” inspired by their experiences as a member of Carnegie Mellon’s class of 2024. The pandemic, Schechter shared, had a massive impact on the campus community, leading much of the class to use Discord as a primary form of communication. The students built an entire language of custom emotes, which developed consistent meanings and patterns of use before becoming words used in their normal speech, as well as in Zoom class chats. 

To explain the “grammar” of emotes, Schechter walked the audience through three different types of emoji use: post-text, included after a message; co-text, used in conjunction with words; and pro-text, which function as words. Each of these usage types has a different effect on how the message is interpreted, offering people a way to modify the tone of their message in ways unachievable by normal text. Schechter also explained how the emotes that the 2024 Tartans created, for instance, “thowo” and “thunj,” have been passed down to the subsequent Carnegie Mellon classes, with each year developing new uses and definitions.

Ribbit: language visualization

The third speaker, Jack Sun, is currently a senior studying linguistics at Carnegie Mellon. As part of his senior thesis, he created a language app named Ribbit, an application that helps learners of Chinese improve their basic vocabulary and tonal pronunciation. 

His inspiration was PRAAT, a 90s-era linguistic analysis interface that facilitates the visualization of speech. Although PRAAT is very detailed, Sun explained, the program is difficult to understand without specialized knowledge. Ribbit makes speech visualization understandable to an average user, involving an audio analysis API that returns raw data extractions as analyzed relative values and feedback, allowing the user to see a visual representation of their pronunciation compared to the ideal. 

The app allows for voice calibration, has a custom pitch floor and ceiling, and fine-tunes the raw audio to remove and fill in any 0Hz values common in Chinese tone 3-4 intonations. 

After a user’s audio sample is analyzed, the app provides specialized feedback to improve the accuracy of the pronunciation. The app itself features a scrollable lilypad map, themed modules, and rewards, including a streak system. Sun is planning to begin user testing for the app soon. 

Working at Duolingo

Estelle Tian, who graduated from Carnegie Mellon in 2016, offered a look into linguistics beyond academia through the blank eyes of Duo the Owl. 

She began at Duolingo as a contractor in 2018, and, after a period of jumping roles, was “rebranded” into a Learning Designer and occasional mascot. She has written a great deal of content types for the app, including grammar explanations and rules for the app to detect mistakes in written questions. 

Now, though, her primary job is to deal with ChatGPT — namely, to fact-check and correct its output. Duolingo, she reported, has been experimenting with AI prompt engineering to aid the creation of new language modules in hopes of speeding up the processing time necessary to make a new course from a decade to just a few years. 

However, this experimentation hasn’t gone smoothly all the time. Tian gave the audience an example of a ChatGPT example for feedback on a Spanish lesson: “Oops, you got the opposite! The word for man is hombre. Mujer is the word for woman. Just remember, hombre sounds like homeboy for man!”

This event was hosted by The Linguistics Club, an on campus organization.