Boli: A dataset for understanding stuttering experience and analyzing stuttered speech

TL;DR

The paper addresses the scarcity of diverse, language-rich stutter datasets in the Indian context by introducing Project Boli, a multilingual, open-access corpus containing read and spontaneous speech, five stutter-type annotations, and extensive demographic and experiential questionnaire data. It documents data collection, manual word-level annotations, and initial analyses, including stutter-type classification using MFCC features with various classifiers and cross-dataset evaluation, along with ASR comparisons (Wav2Vec2.0 and Whisper). The contributions include a publicly available dataset with detailed annotations and cross-domain validation insights that can improve stutter detection and robust ASR in multilingual settings. The work has practical implications for developing accessible voice technologies for people who stutter in India and sets the stage for dataset expansion and broader language coverage.

Abstract

There is a growing need for diverse, high-quality stuttered speech data, particularly in the context of Indian languages. This paper introduces Project Boli, a multi-lingual stuttered speech dataset designed to advance scientific understanding and technology development for individuals who stutter, particularly in India. The dataset constitutes (a) anonymized metadata (gender, age, country, mother tongue) and responses to a questionnaire about how stuttering affects their daily lives, (b) captures both read speech (using the Rainbow Passage) and spontaneous speech (through image description tasks) for each participant and (c) includes detailed annotations of five stutter types: blocks, prolongations, interjections, sound repetitions and word repetitions. We present a comprehensive analysis of the dataset, including the data collection procedure, experience summarization of people who stutter, severity assessment of stuttering events and technical validation of the collected data. The dataset is released as an open access to further speech technology development.

Figures (3)

• Figure 1: Spectrograms illustrating a few stutter types associated with a male speaker: (a) Sound Repetition (SR), (b) Prolongation (PR), and (c) Block (B).
• Figure 2: Proposed methodology for stutter-type classification from stuttered speech signals.
• Figure 3: Most stuttered sounds based on information shared through the questionnaire form (collected from 67 subjects)