Detecting Stuttering with Artificial Intelligence: A Hybrid Method for Brazilian Portuguese

2026-04-12 · Source: Paper Index on ACL Anthology · Field: Health & Wellbeing — Clinical Care & Medical Practice, Medical Devices & Health Technology · Depth: Advanced, quick

Summary

A new software system has been developed to automatically detect and classify stuttering-related disfluencies in Brazilian Portuguese, addressing the traditional manual, subjective, and time-consuming nature of speech-language assessment. The system utilizes a two-stage hybrid approach: initially, deterministic algorithms based on automatic speech recognition (ASR) and temporal data identify simple disfluencies like repetitions and pauses. Subsequently, a hierarchical architecture, combining a Kohonen network (Self-Organizing Map, SOM) and a Multilayer Perceptron (MLP), classifies complex disfluencies such as blocks and prolongations using acoustic features. An initial dataset, annotated by specialists, was created due to the lack of public resources for this task in Brazilian Portuguese. The system achieved an 89.5% accuracy in classifying complex disfluencies, with a Matthews Correlation Coefficient (MCC) of 0.812.

Key takeaway

For speech-language pathologists assessing stuttering in Brazilian Portuguese, this AI-driven system offers a promising tool to enhance objectivity and efficiency. Its 89.5% accuracy in complex disfluency classification suggests it can significantly support clinical decision-making, reducing manual effort and standardizing assessments. Consider integrating such automated tools to streamline diagnostic processes and establish baselines for patient progress.

Key insights

A hybrid AI system automates stuttering detection and classification in Brazilian Portuguese with high accuracy.

Principles

• Combine deterministic and AI methods for robust detection.
• Specialist-annotated data is crucial for new language tasks.

Method

The method involves a two-stage hybrid approach: ASR and temporal analysis for simple disfluencies, followed by a hierarchical SOM-MLP architecture using acoustic features for complex disfluency classification.

In practice

• Use ASR for initial disfluency identification.
• Employ SOM-MLP for complex acoustic pattern classification.

Topics

• Stuttering Detection
• Brazilian Portuguese
• Automatic Speech Recognition
• Kohonen Network
• Multilayer Perceptron

Best for: NLP Engineer, AI Scientist, Research Scientist, AI Engineer, Domain Expert

Related on AIssential

• Automatic Speech Recognition for Child Reading: A Phonemic Approach using Isolated Words in Brazilian Portuguese — A phonemic approach with fine-tuned Wav2Vec2.0 significantly improves ASR for child reading assessment in Brazilian Portuguese.
• Segment-Level Mandarin Chinese Speech-Based Cognitive Impairment Detection via an Autoencoder with Contrastive Learning — Segment-level speech representation with autoencoders and contrastive learning improves cognitive impairment detection in low-resource settings.
• Systematic Study of Dysarthric Speech Recognition: Spectral Features and Acoustic Models — Incorporating Pitch features and optimizing training frame overlap significantly improves dysarthric speech recognition with F-TDNN models.
• Parallel hierarchical encoding of linguistic representations in the human auditory cortex and recurrent automatic speech recognition systems — Human brains and recurrent ASR systems employ parallel hierarchical strategies to transform sound into linguistic meaning.
• Software for Automatic Speech Recognition via Whisper models applied to Oral History interviews in the Portuguese language — Ethos AT provides private, accessible automatic speech transcription for Portuguese using Whisper models.
• Corti's new Symphony for Speech-to-Text model beats OpenAI at medical terminology accuracy, highlighting the value of specialized AI — Specialized AI models significantly outperform generalist ones in highly regulated, domain-specific fields like healthcare.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Paper Index on ACL Anthology.