Khmer ASR · Speech Technology · Low-Resource NLP

Natt Korat

Lecturer-Researcher at CADT. Khmer ASR researcher and speech technology specialist.

Lecturer-Researcher at CADT building speech and language technologies for Khmer and other low-resource settings, with work spanning ASR, speech data processing, and multilingual NLP.

Background

About Me

Natt Korat
Natt Korat
Lecturer-Researcher · Speech Technology Unit · CADT, Cambodia

I am a Lecturer-Researcher at the Cambodia Academy of Digital Technology (CADT), where I develop speech and language technologies for low-resource languages. My research focuses on Khmer automatic speech recognition, speech data processing, and multilingual natural language processing.

Education
M.S. in Computer Science
Completed January 2026
B.S. in Computer Science
AGA Institute
2023
B.A. in Psychology
Royal University of Phnom Penh
2022
5
Publications
3
Projects
3
Degrees
🇰🇭
Focus Language
Research Interests
🎙️ Speech Recognition (ASR) 🔬 Self-Supervised Learning ⚡ Parameter-Efficient Fine-Tuning 🌏 Low-Resource Languages 🇰🇭 Khmer NLP 📰 Information Extraction 🏥 AI for Public Health 🔉 Speech Data Processing 🤗 Transformer Models 🌐 Multilingual NLP

Research Output

Publications

Selected publications and workshop contributions on Khmer NLP, information extraction, applied machine learning, and speech-related technologies.

Journal 2025
Khmer News Classification in Low-Resource Settings: A Comparative Analysis of Embedding Methods
Korat, N., Heang, S., & Lay, V.
Journal on Information Technologies & Communications
DOI
Conference 2025
Epidemic Event Extraction from News Media Using Large Language Models
Korat, N., Kak, S., Lay, V., Uraiwan, B., & Waranrach, V.
International Conference on Digital Economy and Fintech Innovation (DEFI 2025)
DOI
Conference 2025
Building a Khmer NER Benchmark from Health News Data Towards Event Extraction
Chiep, C., Korat, N., Lay, V., & Ly, R.
ASEAN Conference on Emerging Technology (ACET 2025)
HAL
Poster 2025
Toward Multilingual Epidemic Event Extraction for Low-Resource Languages
Korat, N., Waranrach, V., & Kak, S.
UEC International Seminar Poster Session
Workshop 2024
Robust and Efficient Recognition of Khmer License Plates Using YOLOv5 and Parseq OCR
Kong, P., Korat, N., & Veng, P.
CV4DC Workshop, Asian Conference on Computer Vision (ACCV 2024)

Work in Progress

Featured Projects

🎙️
Khmer Speech Technology
Khmer ASR systems built on XLS-R 300M with CTC decoding, speaker diarization (pyannote), and INT8 quantization via ONNX Runtime for lower-latency transcription workflows.
Focus: practical Khmer transcription for applied research settings.
Wav2Vec2 / XLS-R CTC Diarization ONNX
🔊
Speech Dataset Pipeline
Pipeline for collecting and processing Khmer speech data from broadcast media and public web sources, with text normalization, forced alignment, and PyCTCDecode + KenLM beam-search decoding.
Focus: repeatable dataset preparation for low-resource ASR experiments.
Data Collection KenLM Forced Alignment
🦠
Event Extraction System
Pipeline for extracting epidemic-related events from multilingual news and social media with transformer models and LLMs, turning unstructured text into structured public health signals.
Focus: multilingual event signals for public health monitoring.
NER LLM Multilingual Event Extraction

Get in Touch

Contact

I welcome research collaborations, academic discussions, and project conversations related to speech technology, Khmer NLP, and low-resource language AI.

Send a Message

Write a short note and your email client will open with a pre-filled draft.

Opens your local mail app (Gmail, Outlook, etc.)