Using computational tools to collect naturalistic data: Phonetic alignment in political debates

Language science research, particularly in spoken language, has often faced challenges when it comes to collecting naturalistic production data. While in-lab production studies specifically focus on word- or sentence-level productions with extractions of certain phonetic features, sociolinguistic studies focus on understanding characteristics of a speaker and how they manifest in their production, and the interaction of these two domains led to many theoretical and empirical advancements in sociophonetic research. However, due to the laborious nature of procuring naturalistic data along with significant logistical and methodological barriers, these data are scarce. Here, we created a brand-new computational pipeline called the Acoustic Analysis Pipeline, which is a collection of various modules that can be used independently or combined to procure, process, and analyze naturalistic, multi-speaker acoustic-phonetic data automatically. Currently, the pipeline’s main components are the detection, transcription, and analysis of vowel pitch and vowel formants in multi-speaker conversations. We present some preliminary data of phonetic alignment from political debates in the USA (2016–2024).