កម្រងទិន្នន័យភាសាខ្មែរKhmer Corpora

About Me

Chem Vatho, PhD

Phonetician and computational linguist: speech and language technology for low-resource languages

Most recent position

CIF Tandem Fellow, IfL-Phonetik, University of Cologne

Aug – Oct 2026

Cologne International Forum Tandem Fellowship, awarded for the project “Revitalizing and Documenting Kölsch: From Archival Materials to an AI-Ready Speech Recognition and Synthesis System for the Endangered Kölsch Dialect.” Tandem partner: Dr. Simon Rössig.

  • Built an end-to-end Kölsch ASR pipeline: OCR digitisation of the Alles Kölsch corpus, MMS forced alignment, phonological normalisation, and Wav2Vec2 XLS-R-300M fine-tuning — 15% WER / 12% CER on a dialect with no standard orthography.
  • Developed a phoneme-level recognition and forced-alignment system (w2v-BERT 2.0, ~63-phone IPA inventory) with TextGrid output, a 7,354-entry pronunciation dictionary, and a rule-based post-G2P normalisation module.
  • Compared XLS-R-300M against MMS-1b full fine-tuning and adapter variants using paired bootstrap significance testing.

Research software and infrastructure

  • XLinCoLab (github.com/chemvatho/XLinCoLab) — reproducible Python analysis pipelines for experimental linguistics across four paradigms: EEG (MNE-Python, N400/P600 ERP extraction, Morlet time–frequency analysis), acoustic analysis (Praat/Parselmouth, librosa, CREPE), visual-world eye-tracking (growth-curve modelling, cluster-based permutation tests) and reaction time (mixed-effects models). MIT-licensed.
  • XColab Multimodal Embedding Pipeline (github.com/chemvatho/multimodal_Embedding) — a shared 256-dimensional embedding space for EEG, speech, eye-tracking and reaction-time data, with per-modality encoders, L2 normalisation for cross-modal cosine comparison, and PCA/t-SNE/UMAP evaluation. Architecture and evaluation harness complete; encoders untrained, establishing the pre-fine-tuning baseline.
  • Kölsch phoneme-recognition system and fine-tuned model, released on GitHub and Hugging Face; Khmer grapheme-to-phoneme converter on Hugging Face Spaces; Khmer IPA pronunciation dictionary (29,204 entries), BiLSTM-CRF word segmenter and 137-hour aligned corpus.

Education

Dr. phil. in Phonetics, University of Cologne

2022 – 2025

Dissertation: Phonetic Reduction in Central Khmer: An Acoustic, Computational, and Linguistic Analysis. Supervisors: Prof. Dr. Martine Grice and PD Dr. Constantijn Kaland.

MA in Linguistics, Royal University of Phnom Penh

2014 – 2017

Thesis (2016): an acoustic study of the vowel system of Standard Khmer.

BA in Computer Science, Institute of Cambodia-International Cooperation

2010 – 2015

BA in Khmer Literature, Royal University of Phnom Penh

2009 – 2013

Academic positions

Doctoral Researcher, IfL-Phonetik, University of Cologne

2021 – 2025

Lecturer, Royal University of Phnom Penh — Khmer Language and Civilization

2016 – 2022

Taught undergraduate courses in phonetics, Khmer linguistics and literature; supervised student projects.

Vice-Rector Assistant, Royal University of Phnom Penh

2013 – 2016

Coordination and fieldwork for the Linguistic Atlas of Cambodia project.

Academic service

Member, Linguistics Committee, Royal Academy of Cambodia

2018 – 2022

Fellowships, grants and awards

  • CIF Tandem Fellowship, Cologne International Forum, University of Cologne (2026–2027).
  • KAAD scholarship UoC — doctoral studies and fieldwork in Cambodia.
  • NSF-funded participation, SEALS 31, Hawaiʻi, USA.
  • KAAD-funded participation, SEALS 32 and ICAAL 11, Chiang Mai.

Research and technical skills

Speech technology
Self-supervised speech models (Wav2Vec2, XLS-R, w2v-BERT 2.0, MMS, Whisper); CTC and LoRA fine-tuning; KenLM / pyctcdecode beam-search decoding; forced alignment (MFA, torchaudio); grapheme-to-phoneme conversion; neural audio codecs for TTS; CRNN-based OCR.
Phonetics
Acoustic analysis (Praat, Parselmouth, librosa); F0 extraction (Praat, pYIN, CREPE), formant tracking, voice quality (jitter, shimmer, HNR); segmental and prosodic annotation; IPA transcription and transcription-convention design; corpus construction from archival and field recordings; speaker and dialect variation.
Experimental and psycholinguistic methods
EEG/ERP processing and analysis (MNE-Python; N400, P600; time–frequency analysis); visual-world eye-tracking (fixation extraction, growth-curve modelling, cluster-based permutation tests); reaction-time and lexical-decision paradigms; multimodal representation learning across these data types.
Programming and statistics
Python (PyTorch, Hugging Face, NumPy, pandas, scikit-learn, statsmodels); mixed-effects and regression modelling; bootstrap significance testing; Bash; Git; Docker and NVIDIA container workflows; CUDA/GPU training on Grace Blackwell hardware.

Conference presentations

2026

  • 11th International Symposium on Dialectology and Geolinguistics (SIDG XI), Philipps-Universität Marburg, Germany, 7–11 September 2026 (presented 9 September).
  • 34th Annual Conference of the International Association for Forensic Phonetics and Acoustics (IAFPA), Trier University, Germany, 26–29 July 2026 (Oral Session III, 27 July).

2025

  • 34th Annual Meeting of the Southeast Asian Linguistics Society (SEALS 34), Bali, Indonesia.
  • “Flows and Currents in Southeast Asia” Graduate Conference, SOAS University of London, United Kingdom, 3–4 July 2025.

2024

  • 33rd Annual Meeting of the Southeast Asian Linguistics Society (SEALS 33), Taipei, Taiwan, 15–17 June 2024.
  • 12th International Conference on Austroasiatic Linguistics (ICAAL 12), Payap University, Chiang Mai, Thailand, 23–25 October 2024.

2023

  • 32nd Annual Meeting of the Southeast Asian Linguistics Society (SEALS 32), Chiang Mai, Thailand, May 2023 (KAAD-funded).
  • 11th International Conference on Austroasiatic Linguistics (ICAAL 11), hosted by Chiang Mai University, Thailand (online), 26–27 October 2023 (KAAD-funded).

2022

  • 31st Annual Meeting of the Southeast Asian Linguistics Society (SEALS 31), University of Hawaiʻi at Mānoa, Honolulu, USA (hybrid), 19–21 May 2022 (NSF-funded).
  • 10th International Conference on Austroasiatic Linguistics (ICAAL 10), hosted by the Living Tongues Institute for Endangered Languages, USA (online), 2022.

2019

  • 19th International Congress of Phonetic Sciences (ICPhS 2019), Melbourne, Australia, 5–9 August 2019.
  • 8th International Conference on Austroasiatic Linguistics (ICAAL 8), Chiang Mai University, Thailand, 29–31 August 2019.

2016

  • 3rd International Conference on Asian Geolinguistics, Phnom Penh, Cambodia.

2015

  • 6th International Conference on Austroasiatic Linguistics (ICAAL 6), Royal Academy of Cambodia and Center for Khmer Studies, Siem Reap, Cambodia.

Languages

  • Khmer (native)
  • English (fluent)
  • German (good working proficiency, actively developing)
  • Thai (basic)

Publications

Selected

  • An acoustic study of the Standard Khmer vowel system (CJBAR, 2020).
  • An Illustration of the IPA for Phnom Penh Khmer (Journal of the International Phonetic Association, under revision).
  • A Khmer forced-alignment paper (Language Resources and Evaluation, under revision).
  • A monograph based on the dissertation, submitted to Language Science Press.

Conference presentations at ICPhS, IAFPA, SIDG, SEALS, and ICAAL are listed above.