Visualizing multilingual speech articulation: A corpus-based approach

This presentation introduces the SNU Multilingual Articulatory Corpus (SNU-MAC), a large-scale multimodal dataset designed to support research on L1 and L2 speech articulation. The corpus includes synchronized ultrasound tongue imaging, front and lateral lip videos, audio recordings, and detailed language background information. Data include speech samples in Korean, English, Spanish, and Chinese, produced by speakers from diverse linguistic backgrounds, including both native and nonnative speakers. Speech tasks range from picture description and narrative reading to controlled word and syllable production, allowing for analysis across different speaking styles and proficiency levels. The corpus enables close examination of articulatory variation by speakers’ native language and proficiency, such as articulatory consistency across languages in multilingual speakers and L1 influence at the articulatory level. Multimodal data further allow for the analysis of potential mismatches between articulation and acoustics, due to silent articulation or gestural overlap. Sample video clips and data access are available through the project website. Ultimately, the SNU-MAC aims to provide open, structured access to multilingual articulatory data for researchers, educators, and students in phonetics, speech science, multilingualism, and language acquisition. Work supported by SNU Creative-Pioneering Researchers Program.