HUMAN COMPUTER INTERACTION LAB
Design for change and Inclusion
Instagram
Instagram
YouTube
YouTube
X
X
Facebook
Facebook
SELECTED PUBLICATIONS
Advancing automatic speech recognition for low-resource ghanaian languages: Audio datasets for Akan, Ewe, Dagbani, Dagaare, and Ikposo Redirecting
Advancing automatic speech recognition for low-resource ghanaian languages: Audio datasets for Akan, Ewe, Dagbani, Dagaare, and Ikposo Redirecting
Benchmarking Akan ASR Models Across Domain-Specific Datasets: A Comparative Evaluation of Performance, Scalability, and Adaptability
Benchmarking Akan ASR Models Across Domain-Specific Datasets: A Comparative Evaluation of Performance, Scalability, and Adaptability
Abstract page for arXiv paper 2507.02407: Benchmarking Akan ASR Models Across Domain-Specific Datasets: A Comparative Evaluation of Performance, Scalability, and Adaptability
A Cookbook for Community-driven Data Collection of Impaired Speech in LowResource Languages
A Cookbook for Community-driven Data Collection of Impaired Speech in LowResource Languages
This study presents an approach for collecting speech samples to build Automatic Speech Recognition (ASR) models for impaired speech, particularly, low-resource languages. It aims to democratize ASR technology and data collection by developing a "cookbook" of best practices and training for community-driven data collection and ASR model building. As a proof-of-concept, this study curated the first open-source dataset of impaired speech in Akan: a widely spoken indigenous language in Ghana. The study involved participants from diverse backgrounds with speech impairments. The resulting dataset, along with the cookbook and open-source tools, are publicly available to enable researchers and practitioners to create inclusive ASR technologies tailored to the unique needs of speech impaired individuals. In addition, this study presents the initial results of fine-tuning open-source ASR models to better recognize impaired speech in Akan.
Comparative evaluation of learning technologies using a randomized controlled trial: Virtual reality, augmented reality, online video platforms, and traditional classroom learning - Education and Information Technologies
Comparative evaluation of learning technologies using a randomized controlled trial: Virtual reality, augmented reality, online video platforms, and traditional classroom learning - Education and Information Technologies
Learning Satisfaction in Virtual Reality: The Role of Persuasive Design
Learning Satisfaction in Virtual Reality: The Role of Persuasive Design
PROJECTS
Text-to-Speech Voices for African Languages - Funded by Google
Text-to-Speech Voices for African Languages - Funded by Google
Text-to-Speech Interface and Evaluation System
Tɛkyerɛma Pa Project - Funded by Google (Google gift)
Tɛkyerɛma Pa Project - Funded by Google (Google gift)
Tɛkyerɛma Pa Hackathon 2025- Funded by GDI-Hub, UCL
Tɛkyerɛma Pa Hackathon 2025- Funded by GDI-Hub, UCL
Join us in creating innovative solutions for individuals with disabilities. Participate in our annual hackathon that makes a real difference in Ghana and beyond.
UGSpeechData - Funded by Google
UGSpeechData - Funded by Google
The UGSpeechData is a collection of audio speech data of Akan, Ewe, Dagaare, Dagbani, and Ikposo. These languages are among the most spoken languages in Ghana. The uploaded dataset contains a total of 970148 audio files (5384.28 hours) and 93262 transcribed audio files (518 hours). The audio files are descriptions of 1000 culturally relevant images collected from indigenous speakers of each of the languages. Each audio is between 15 to 30 seconds long. More specifically, the dataset contains five subfolders for each of the five respective languages. Each language has at least 1000 hours of speech data and 100 hours of transcribed speech data. Fig. 1 provides details of the transcribed audio corpus, including gender and recording environments for each language.Fig. 1. Details of transcribed audio files
PRODUCTS
Akan to English Translator
Akan to English Translator
Generated by create next app
Keyboards
Keyboards
This repository contains keyboards for 13 African languages - HCI-LAB-UGSPEECHDATA/SPEECH-DATA-KEYBOARDS
Akan ASR
Akan ASR
Transcribe audio files using our models.
Akan TTS
Akan TTS
Text-to-Speech Interface and Evaluation System
AKAN SPELL CHECKER
AKAN SPELL CHECKER
Text-to-Speech Interface and Evaluation System
SOCIAL MEDIA HANDLES
Website
Website
LinkedIn
LinkedIn
DCS HCI Lab | 1,078 followers on LinkedIn. We research & design persuasive, inclusive, & accessible tech for low-resource environments. https://linktr.ee/DCSHCILAB | The DCS HCI Lab is dedicated to advancing the field of human-computer interaction through cutting-edge research interests persuasive technology, extended reality, artificial intelligence, and speech technologies. We have developed various persuasive applications in extended reality for education, fitness, road safety, stress & anxiety. Our mission is to design and develop culturally relevant persuasive, accessible, inclusive, and impactful technological solutions that leverages AI to address real-world challenges in low-resource settings.
HuggingFace
HuggingFace
Persuasive Technology, ASR, LLM, Virtual Reality, Machine Learning and AI
Instagram
Instagram
YouTube
YouTube
GitHub
GitHub
The Human-Computer Interaction Lab in the Department of Computer Science, University of Ghana with interest in LLM, ASR, TTS, AR, and VR applications. - DCS HCI LAB (UG Computer Science) Ghana
UG | Department of Computer Science
UG | Department of Computer Science
X
X
Facebook
Facebook
View on mobile
