EEG studies and machine learning, the tools of language studies pioneers.
When asked “Do you speak another language?” typical responses range from “just a little” and “I speak some” to “I’m pretty good.” But how does one measure the difference between a little, some, pretty good, and fluent? Judging our own expertise in language can be really difficult—not just because it’s difficult to evaluate our own skills but also because there are so many nuances to language. That’s where research can make a significant impact.
Professor Matthew Wilcox (Language Assessment Project Manager, Center for Language Studies) uses machine learning to test an individual’s language abilities, while Professor Ellen Knell (Associate Director for Curriculum and Instruction, Center for Language Studies) hooks her subjects up to an EEG machine to figure out if they have language knowledge that they might not even be aware of. These two ongoing studies in the College of Humanities aim to help participants determine individual proficiency in a foreign language.
Prototype Language Assessment through Machine Learning
Wilcox heads the team dedicated to accurately placing students in language classes. His strategy involves a machine learning system called the Language Ability Self-Evaluation Resource (LASER). LASER is a test that utilizes machine learning to help students determine which language courses fit their skill sets.
LASER consists of two parts. The first half involves gathering students’ language background. This background information establishes a baseline for the type of questions LASER will ask to determine your ability level.
The second half is a written segment that applies the skills the participant already has by asking them to answer questions in the target language. Using this written segment, Wilcox’s team programmed a machine that can look for patterns to determine skill level. This machine learning targets common elements of skill, such as word count, type-token ratio, mean length of utterance, fluency metrics, number of pauses, and more. LASER takes these elements directly from the written portion and calculates them immediately.
Wilcox designed LASER to help foster lifelong language learning. He says, “It’s not a standard test; it’s for you to give your typical ability, help you find the gaps, and then for instructors to be able to use that to improve student learning.”
Interested in giving LASER a try? Go to LASER to measure your skill level.
Neurolinguistic Analysis of Early Second Language Learning
Second language learners often form connections in the brain between foreign words and definitions long before they recognize that connection or have the ability to vocalize it. Since language learners aren’t aware of all the knowledge they have, it’s hard for researchers to measure when exactly language learning begins. Knell and Assistant Professor Jeffrey Green (Linguistics) discussed this hidden connection and determined to conduct a study measuring brainwaves of Chinese learners in the fall of 2022. This will be a replication study based on research done in France that looked at early brainwave signals for French learners.
Their study aims to measure the early neurological connections that form when a student has begun learning a second language. Participants will be attached to an electroencephalography machine (EEG) and monitored with functional near-infrared spectroscopy (fNIRS). These machines will measure brainwaves and brain oxygen levels. Each participant will be shown a series of Chinese words, then identify if they think each word is real or made up. Knell hopes that, like in the original study, the brainwaves will move in specific ways when shown a real word and move differently when presented with a false word. This will be an indicator of early language learning.
All faculty members involved feel that this testing could have important implications for language learning. The results of this study could impact how second-language courses are taught and could lead to future studies about what methods are most effective for learning a second language. While Knell’s and Wilcox’s studies differ in methodology, both of their research results will help pave the way to better understanding language proficiency in students.