Automatic speaking assessment
End-to-end scoring for spoken responses using compact speech encoders and lightweight aggregation.
Research group
Speech Recognition
Recent output
2023-2026
Aalto-listed papers
13
Mobile apps
3 learner tools
Profile
Research that turns speech models into learner-facing systems.
My work connects automatic speech recognition, spoken language assessment, and computer-assisted language learning. The common thread is usable feedback: systems that can help learners, teachers, and researchers even when task-specific data is limited.
Current applied work covers Finland Swedish mispronunciation detection, Whisper-based automatic speaking assessment, Finnish L2 content feedback, and mobile learning apps such as SaySuomi, SaySvenska, and Pop2TalkNordic.
Research
Core directions
A focused research agenda across speech modeling, feedback design, and language learning applications.
Pronunciation feedback
Mispronunciation detection and feedback methods for learners in low-resource language settings.
Learning applications
Speech-enabled mobile tools for pronunciation practice, task completion, and self-study feedback.
Low-resource speech AI
Robust modeling strategies for Finnish, Finland Swedish, and other under-supported tasks.
Publications
Selected work
SLaTE 2025
One Whisper to Grade Them All
Efficient holistic automatic speaking assessment with a single Whisper-small encoder.
Interspeech 2025
Mispronunciation Detection Without L2 Pronunciation Dataset in Low-Resource Setting
A practical case study for Finland Swedish pronunciation feedback with minimal L2 data.
International Journal of Language & Communication Disorders
Self-Supervised App-Based Speech Training for Children With Speech Sound Disorder
SLaTE 2025
Exploring Ordinal Classification for Spoken Language Assessment
Language Testing
Enhancing Second Language Speaking Assessment with Large Language Models
Interspeech 2024
CaptainA Self-Study Mobile App for Practising Speaking: Task Completion Assessment and Feedback with Generative AI
Interspeech 2024
Automated Content Assessment and Feedback for Finnish L2 Learners
IEEE Access
Developing an AI-Assisted Low-Resource Spoken Language Learning App for Children
NoDaLiDa 2023
CaptainA - A Mobile App for Practising Finnish Pronunciation
Mobile Apps
Learner-facing tools built from speech research.
These apps are the applied side of my research: mobile experiences for practicing speaking, pronunciation, and task completion with speech-driven feedback.
SaySuomi
A mobile practice app for Finnish learners, connecting speaking tasks with pronunciation and content-oriented feedback.
SaySvenska
A mobile app focused on Swedish and Finland Swedish pronunciation practice, aligned with low-resource mispronunciation detection research.
Pop2TalkNordic
A Nordic language-learning app that frames pronunciation and speaking practice through interactive, learner-friendly mobile activities.
Video Demos
Short demos of applied speech-learning systems.
These videos show the learner-facing side of my work, from mobile speech practice to automatic content assessment and personalized feedback.
YouTube / Interspeech 2024 Show & Tell
Automatic content assessment and personalized feedback
Demo accompanying CaptainA self-study mobile app for practising speaking: task completion assessment and feedback with generative AI, showing how spoken answers are assessed for content and turned into learner feedback.
Watch on YouTubeZenodo / Video recording
Pop2TalkNordic mobile app for the Swedish speech sound disorder study
Video demo of the Pop2TalkNordic mobile app used in the Swedish speech sound disorder study, the app behind the self-supervised app-based speech training work.
Phan Chi, N. (2026, February 8). Video demo of the Pop2TalkNordic mobile app for the Swedish speech sound disorder study [Video recording]. Zenodo. doi:10.5281/zenodo.18526178
Watch on ZenodoExpertise
Speech assessment and learner feedback.
Speech AI methods
- Automatic speech recognition
- Speech representation learning
- Automatic speaking assessment
- Mispronunciation detection
- Large language models for assessment
Learning feedback
- Pronunciation feedback design
- Task completion assessment
- Mobile speaking practice
- Learner-facing feedback loops
- App-based speech training
Research contexts
- Finnish L2 learning
- Finland Swedish pronunciation
- Nordic language learning
- Speech sound disorder training
- Educational assessment
Contact