This week I’m in Philadelphia at annual meetings of the Linguistic Society of America and the American Dialect Society. I gave three talks, which you can download here, I officially started two new service positions, and I got an award! It was a busy conference for me!
Saturday morning’s ADS talk on Georgia Prelaterals
Download the slides here!
On Saturday morning at the ADS session, Peggy Renwick presented research on behalf of our research term on prelateral front vowels and how they’ve changed over time in Georgia. The talk is called “Complementary pre-lateral mergers across ethnicities and generations in Georgia.” Overall, we find that the various prelateral mergers peaked and reversed at different times for Black and White speakers, and for the two feel-fill and fail-fell pairs.
Saturday afternon’s ADS poster on Utah English
Download the poster here!
On Saturday afternoon, I presented with my student, Hallie Davidson, a poster called “Variation in Early 20th Century Utah English.” This is the first time I’ve presented results from a collection of oral narratives I acquired called the “Kohler Tapes.” I’ve written blog posts about how I acquired the tapes, the process of digitizing them, and a summary of the metadata, so I won’t repeat that here. But, because Hallie has been working on transcribing these tapes, we finally have some really preliminary linguistic results to present. They’re just impressionistic right now, nothing really conclusive, but we show that there’s a lot of variation in this collection. We’re excitedly working on some developments that involve training an AI model on this data so we can get the rest transcribed and we look forward to doing acoustic analysis soon!
Sunday morning’s LSA talk source separation
Download the slides here!
On Sunday morning, my final presentation was with Lisa Morgan Johnson (who presented with me) and Earl Kjar Brown. We were interested in source separation, which is where you separate overlapping audio into different tracks, one per speaker. We wanted to see if current algorithms produced reliable enough data for sociophonetic analysis. We took two recordings and processed them using typical methods and called that output the ground truth. We than synthetically overlapped the audio to create a cacophonous overlapping recording. We then tried three source separate models and processed the resulting audio files and compared them to the baseline.
We’re cautiously optemisitic about the results. There were some differences here and there, but on the whole the output formant measurements were pretty similar. However, we did notice that while the mean formants per vowel per speaker were similar, at the indidual token level, there were larger differences.
Anyway, we hope this will spark some interest in source separation and maybe help people recover audio that was previously unanalyzable!
Other Things
At the ADS Business Meeting, I was voted a member of the American Dialect Society Executive Committee! I replace Alicia Wassink, who was just voted President-Elect of the Linguistic Society of America.
At that same meeting, I was also voted a Kurath Fund Trustee, which is a position within the Linguistic Atlas Project. I’ll serve with Jennifer Cramer and Dennis Preston.
Finally, at the ADS Breakfast on Saturday morning, I received the 9th Audrey Duckert Memorial Travel Award to attend these ADS meetings.
I’m so humbled to be considered for these positions and award! Thank you to all those who were involved!