Ph.D. candidate Anna Seo Gyeong Choi has had one paper and two posters accepted for presentation at upcoming summer conferences:

FAccT 2023 - Association for Computing Machinery Conference on Fairness, Accountability, and Transparency (ACM FAccT), Chicago, USA, June 12-15, 2023

• (Talk) Augmented Datasheets for Speech Datasets and Ethical Decision-Making.”, Papakyriakopoulos, Orestis*, Anna Seo Gyeong Choi*, William Thong, Dora Zhao, Jerone Andrews, Rebecca Bourke, Alice Xiang, Allison Koenecke. 2023. “ (* equal contribution co-first author)

• (Poster) - Augmented Datasheets for Speech Datasets and Ethical Decision-Making. Papakyriakopoulos, Orestis*, Anna Seo Gyeong Choi*, Alice Xiang, Allison Koenecke. 2023.   (* equal contribution co-first author)

IC2S2 2023 - the 9th International Conference on Computational Social Science (ICS2), Copenhagen, Denmark, July 17-20, 2023

• (Poster) - Auditing Korean Speech Datasets for Dialectal Fairness in Speech-to-Text Applications,  Anna Seo Gyeong Choi, Allison Koenecke. (non-archival)

Seven Phonetics Lab researchers will be presenting nine papers and posters at  ICPhS 2023  (the 20th International Congress of Phonetic Sciences), which will be held August 7-11, 2023 in Prague, the Czech Republic.

The researchers and the paper/poster titles are:  

Chloe Kwon, Sam Tilsen, and John Whitman:

• Morpho-phonological effects on the phonetic characteristics of tense consonants in Korean compounds

Francesco Burroni:

• Lexical tones are times to articulatory gestures

Seung-Eun Kim:

•  Evidence for preplanned and adaptive F0 control (Seung-Eun Kim) 

Seung-Eun Kim & Sam Tilsen:

•  Investigations of F0 control: pitch targets vs. pitch register (poster)

Simon Roessig:

• Prosody of pre-focal background depends on following focus. 

• Pagel, L., M. Sóskuthy, S. Roessig & D. Mücke.    A kinematic analysis of visual prosody: head movements in habitual and loud speech. 

• Parrell, B., A. Mefferd, S. Harper, S. Roessig & D. Mücke.   Using computational models to characterize the role of motor noise in speech: the case of amyotrophic lateral sclerosis. 

• Lorenzen, J., S. Roessig & S. Baumann.  Redundancy and individual variability in the prosodic marking of information status in German. 

Sireemas Maspong

• Voice quality is not an obligatory stage in tonogenesis: A case study of Eastern Khmu

Phonetics Lab Alumnus Dr. Masa Gibson recently created a new music video titled "Glint of Green" for the 2023 WNYC Public Song Project.  The Project explores the creative use of songs, books, and movies that have entered the US public domain.

Masa's composition is an original song for voice and glass bottles, composed and performed by Masa where they sing the poem "Young Witches" by Marian Thanhouser (1927). Masa's composition is unique in that the musical accompaniment is performed exclusively on glass bottles, images of which are featured in this video. 

Masa's "Glint of Green" was one of 80 compositions submitted the Public Song Project, and it showcases Masa's broad talents in singing, acting, music composition, and film production, which are spare-time pursuits that complement their speech synthesis R&D for Synfonica LLC.  

Masa commented that: "My interest/knowledge in acoustics definitely played a hand in the project", and they had to review their knowledge of Helmholtz resonators before using the bottles as musical instruments.

Phonetics Lab alumnus Dr. Linda Heimisdottir (left, in the flowered dress) recently met with the Iceland President HE Guðni Th. Jóhannesson (center)  as part of a new project to use OpenAI's GPT-4 to preserve the Icelandic language. 

 

This project - initiated during a visit by President Jóhannesson and a language technology delegation to Open AI’s headquarters last spring -  is a partnership between the Icelandic government, Icelandic NLP/AI company Miðeind, and OpenAI, where Miðeind is working with OpenAI to train and fine-tune GPT-4 on Icelandic.  Representatives of OpenAI (including Anna Makanju, their Head of Public Policy) participated in the meeting, and they also attended a local symposium about AI and low-resource languages. 

 

The partnership was envisioned not only as a way to boost GPT-4’s ability to service a new corner of the world, but also as a step towards creating resources that could help preserve other low-resource languages.