SEP-28k: A Dataset for Stuttering Event Detection From Podcasts With People Who Stutter
This addresses the need for better tools to track fluency for speech pathologists and improve speech recognition for people with atypical speech, though it is incremental as it primarily expands existing data resources.
The authors tackled the problem of limited data for stuttering event detection by introducing SEP-28k, a dataset with over 28k labeled clips from podcasts featuring people who stutter, and showed that increasing training data improved detection performance by 28% and 24% F1 on benchmarks.
The ability to automatically detect stuttering events in speech could help speech pathologists track an individual's fluency over time or help improve speech recognition systems for people with atypical speech patterns. Despite increasing interest in this area, existing public datasets are too small to build generalizable dysfluency detection systems and lack sufficient annotations. In this work, we introduce Stuttering Events in Podcasts (SEP-28k), a dataset containing over 28k clips labeled with five event types including blocks, prolongations, sound repetitions, word repetitions, and interjections. Audio comes from public podcasts largely consisting of people who stutter interviewing other people who stutter. We benchmark a set of acoustic models on SEP-28k and the public FluencyBank dataset and highlight how simply increasing the amount of training data improves relative detection performance by 28\% and 24\% F1 on each. Annotations from over 32k clips across both datasets will be publicly released.