Luganda Mental Health Dialogues: Simulated Counselling Sessions

A conversational speech dataset of simulated mental health counselling sessions in Luganda, recorded in Uganda. It features dialogues between a Helper (counsellor) and a Seeker (client) discussing mental health topics. The dataset is designed for research in automatic speech recognition, speaker diarization, and speaker role classification for a low-resource African language.

Use Cases

• Train an automatic speech recognition (ASR) model on Luganda audio from simulated counselling sessions.
• Perform speaker diarization to segment audio by the Helper and Seeker speaker roles.
• Build a classifier to identify the speaker role (Helper vs. Seeker) within each dialogue.
• Conduct gender classification analysis using the audio from the two speakers in each session.

Strengths

• Contains simulated dialogues for a low-resource African language (Luganda).
• Designed for multiple speech tasks including ASR, diarization, and role classification.
• Features two distinct speaker roles (Helper and Seeker) per conversation.

Limitations

• Dataset size, row count, and audio duration are unknown.
• Data is simulated, not from real clinical sessions, which may limit ecological validity.
• Specifics on recording conditions, audio quality, and speaker demographics are not provided.

Provenance

Source

huggingface

Collection Method

Simulated recordings of mental health counselling sessions.

Freshness

Last updated 2026-03-16.

Geography

Uganda