The Bluegrass corpus: Audio-visual stimuli to investigate foreign accents

Abstract

The Bluegrass corpus includes sentences from 40 pairs of speakers. Participants from the Bluegrass Region rated one speaker from each pair as having a native North American English accent and the other as having a foreign accent (Experiment 1). Furthermore, speakers within each pair looked very similar in appearance, in that participants rated them similarly likely to speak with a foreign accent (Experiment 2). For each speaker we selected eight sentences based on participants’ ratings of difficulty (Experiment 3). The final corpus includes a selection of 640 sentences (80 speakers, 8 stimuli per speaker) freely available through the Open Science Framework. Each sentence can be downloaded in different formats (text, audio, video) so researchers can investigate how audio-visual information influences language processing. Researchers can contribute to the corpus by validating the stimuli with new populations, selecting additional sentences, or finding new TED videos featuring appropriate speakers to answer their research questions.