Active Projects
Student-led research is the beating heart of the lab. Read on to discover what students are uncovering as junior researchers in CSL.
Acoustics of autistic stuttering
Did you know that stuttering is a common feature of autistic speech, and that the acoustic properties of austistic stuttering are different from neurotypical stuttering? We are conducting a corpus analysis of the acoustics of stuttering in a corpus of Dutch childrens' speech.
Online disfluency prediction
One big challenge for technology like automatic speech recognition is that when tools are used in an 'on-line' or live setting the system doesn't know what's coming next in speech, which would otherwise be important contextual information. This is especially useful in limiting corrections to disfluent (i.e. mistake-laden) speech. We are investigating what acoustic cues can predict future disfluent speech behavior to ameliorate this problem and improve ASR performance.
[t]-glottalization in Utah English
How do you pronounce the 't' in words like mountain or curtain? If it feels like the [t] is in the back of your throat instead of the tip of your tongue, then your speech has t-glottalization, which is a common feature of North American English. We are investigating how surprisal, a mathematical measure of how important a word is in context, impacts the speech of speakers for whom t-glottalization is a salient dialect feature (even if it isn't unique). This work is in collaboration with University of Washington in Washington State and Brigham Young University in Utah.
Phonogramics and Cannery Labour
The racialization of Chinese immigration to the Pacific Northwest during the salmon boom of the late 19th and early 20th centuries has left local historians at a loss. Where records of immigrant labourers exist, they are often simply labeled as 'Chinamen' or 'China labour.' In collaboration with the Washington State Archives and the Center for Pacific Northwest Studies, and supported by the James Scott Fellowship, we uncover the available name records from the archives of two PNW canneries. We use phonogramics to map migration patterns from China to Washington State during this formative era.
Folk Linguistics of AI
Emoji at the beginning of subtitles? Em-dash? Rule of three? If these ideas are familiar, you've probably heard some folk linguistic observations about what makes AI-generated text different from human-generated text. These folk ideas are patterns that humans can see at a glance that differ from the features that auutomatic AI detectors use. We are investigating how useful these folk ideas are at actually classifying text, and categorizing these features according to their linguistic complexity.