Events

The use of corpora in acoustic analyses has become a standard practice in phonetic phonological research, offering high ecological validity (see e.g. Beckman, 1997; Warner, 2012; Tucker & Mukai, 2023 for a discussion on validity). However, compiling corpora and looking for specific phenomena can be time and resource-consuming. In response to this challenge, we developed a program named Pyrlato, which we aim to demonstrate. Pyrlato is a novel tool designed for creating corpora of real-world spoken data from the web. The tool extracts audio files from YouTube, cutting and extracting desired segments such as specific phonemes, syllables, or words found in YouTube videos. This enables the creation of corpora with tens of thousands of tokens within a few computational hours. Pyrlato works across Dutch, English, French, German, Indonesian, Italian, Japanese, Korean, Portuguese, Russian, Spanish, Turkish, Ukrainian, and Vietnamese, i.e. those languages for which YouTube provides automatic subtitles. The software searches for the desired string in the subtitles and, upon finding the match, extracts the relevant audio extract containing the string in .mp3 format (other formats are also possible).

The demonstration will showcase Pyrlato's online version and the application of some case studies.

• Beckman, M.E. (1997).A typology of spontaneous speech. In Y. Sagisaka, N. Campbell, & N. Higuchi (Eds.), Computing Prosody: Computational Models for Processing Spontaneous Speech (pp. 7–26). Springer. http://dx.doi.org/10.1007/978-1-4612-2258-3_2.
• Tucker, B.V., & Mukai, Y. (2023). Spontaneous speech. Cambridge University Press. http://doi.org/10.1017/9781108943024.
• Warner, N. (2012). Methods for studying spontaneous speech. In A. Cohn, C. Fougeron, & M. Huffman (Eds.), The Oxford Handbook of Laboratory Phonology (pp. 621–633). Oxford University Press.