--- title: "MekongPhon: A Large-Scale Parallel IPA Corpus — Shurtz 2026" description: "Grapheme-to-phoneme (G2P) conversion for Lao and Khmer has lacked reliable tools." authors: ["Ammon Shurtz", "Christian Richardson", "Stephen D. Richardson"] year: 2026 venue: "Proceedings of the Fifteenth Language Resources and Evaluation Conference" source: "https://aclanthology.org/2026.lrec-1.129/" doi: "10.63317/4bb9rvdshu4z" licence: "CC BY-NC 4.0 (ELRA, LREC proceedings)" licence_url: "https://creativecommons.org/licenses/by-nc/4.0/" summary_licence: "Elamur Cosmic License" canonical: "https://research.elamur.ai/free/en/papers/shurtz-2026-mekongphon-large-scale-parallel-ipa/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/shurtz-2026-mekongphon-large-scale-parallel-ipa/) [home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/6/) / MekongPhon: A Large-Scale Parallel IPA Corpus for Lao and Khmer # MekongPhon: A Large-Scale Parallel IPA Corpus for Lao and Khmer **Authors:** [Ammon Shurtz](https://md.research.elamur.ai/free/en/authors/ammon-shurtz/), Christian Richardson, [Stephen D. Richardson](https://md.research.elamur.ai/free/en/authors/stephen-d-richardson/) · **Year:** 2026 · **Venue:** Proceedings of the Fifteenth Language Resources and Evaluation Conference · **Licence:** [CC BY-NC 4.0 (ELRA, LREC proceedings)](https://creativecommons.org/licenses/by-nc/4.0/) **Links:** [Publisher's page](https://aclanthology.org/2026.lrec-1.129/) · [PDF](https://aclanthology.org/2026.lrec-1.129.pdf) · [DOI 10.63317/4bb9rvdshu4z](https://doi.org/10.63317/4bb9rvdshu4z) · [OpenAlex](https://openalex.org/W7160029754) **Citations:** 0 citations (openalex-snapshot · 2026-10-14) ## Summary Grapheme-to-phoneme (G2P) conversion for Lao and Khmer has lacked reliable tools. On the authors' gold test data, the widely used Epitran scores a CER of 105.57 for Khmer, because its mapping file leaves out all Khmer vowel signs, and 33.48 for Lao. To address this, the authors build MekongPhon. They scrape human-transcribed IPA entries from two linguist-compiled dictionaries in the SEALang Library to form a seed lexicon: Headley (1977) for Khmer and Kerr (1972) for Lao. They store the lexicon in a trie and run greedy longest-match conversion over millions of OPUS sentences from NLLB, ParaCrawl and OpenSubtitles. Any sentence with leftover Khmer or Lao script characters is discarded, so precision takes priority over coverage. The resulting orthography–IPA parallel data is used to train Transformer sequence-to-sequence G2P models, which reach a CER below 2% on held-out tests. The authors also release fast, linguistically informed rule-based scripts that beat Epitran by 6–71 CER points. Because the corpus, lexicons, scripts and models are all public, the work gives speech and phonology research on these underserved languages a solid base and offers a recipe for other low-resource scripts. ## Key points - **Corpus size:** MekongPhon contains 1,314,779 Khmer and 367,388 Lao sentence pairs. These were filtered from 8,962,874 Khmer and 4,609,363 Lao unique OPUS sentences. - **Seed lexicon:** It holds 8,026 Khmer and 6,039 Lao word–IPA pairs, supplemented by hand-added numerals and punctuation. The Lao entries keep both tonal and non-tonal IPA. - **Lexicon quality:** Two authors manually checked 500 random entries per language. Accuracy was 99.2% for Khmer and 97.2% for Lao. Typical errors were final consonants given an inherent vowel (Khmer) and wrong vowels or finals (Lao). - **Model results:** The Transformer G2P models stay under 2% CER on held-out sets. The rule-based tools trade some accuracy for speed but still clearly outperform Epitran. - **Availability and limitations:** All resources are released at github.com/byu-matrix-lab/MekongPhon. Noted limitations include: - Lao tones follow the Vientiane dialect. - The OPUS source data is noisy. - The neural models are sensitive to errors in the lexicon. *Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.* Fields: [Linguistics & languages](https://md.research.elamur.ai/free/en/shelves/linguistics/) ## Cite The paper itself, not this page. BibTeX from the library's record of the work. ```bibtex @inproceedings{shurtz2026mekongphon, title = {{MekongPhon: A Large-Scale Parallel IPA Corpus for Lao and Khmer}}, author = {Shurtz, Ammon and Richardson, Christian and Richardson, Stephen D.}, booktitle = {Proceedings of the Fifteenth Language Resources and Evaluation Conference}, year = {2026}, doi = {10.63317/4bb9rvdshu4z}, url = {https://doi.org/10.63317/4bb9rvdshu4z} } ``` - [← previous MedAraBench: Large-Scale Arabic Medical Question Answering Dataset and Benchmark](https://md.research.elamur.ai/free/en/papers/abu-2026-daoud-medarabench-large-scale-arabic-medical/) - [next → Memorization or Lucky Guesses: Detecting Short Sequences from Copyrighted Dutch News in LLM Output](https://md.research.elamur.ai/free/en/papers/veerbeek-2026-memorization-or-lucky-guesses-detecting/) - [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/6/) ## References 16 references, 3 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet. 1. Shiyang Cheng, Pengcheng Zhu, Jueting Liu, and Zehua Wang. 2024. A survey of grapheme-to-phoneme conversion methods. Applied Sciences, 14(24):11790. [find](https://scholar.google.com/scholar?q=Shiyang+Cheng%2C+Pengcheng+Zhu%2C+Jueting+Liu%2C+and+Zehua+Wang%2E+2024%2E+A+survey+of+grapheme%2Dto%2Dphoneme+conversion+methods%2E+Applied+Sciences%2C+14%2824%29%3A11790%2E) 2. Alexis Conneau, Kartikay Khandelwal, Naman Goyal, Vishrav Chaudhary, Guillaume Wenzek, Francisco Guzmán, Édouard Grave, Myle Ott, Luke Zettlemoyer, and Veselin Stoyanov. 2020. Unsupervised cross-lingual representation learning at scale. In Proceedings of the 58th Annual Meeting of the Association fo… [find](https://scholar.google.com/scholar?q=Alexis+Conneau%2C+Kartikay+Khandelwal%2C+Naman+Goyal%2C+Vishrav+Chaudhary%2C+Guillaume+Wenzek%2C+Francisco+Guzm%C3%A1n%2C+%C3%89douard+Grave%2C+Myle+Ott%2C+Luke+Zettlemoyer%2C+and+Veselin+Stoyanov%2E+2020%2E+Unsupervised+cross%2Dlingu) 3. Angela Fan, Shruti Bhosale, Holger Schwenk, Zhiyi Ma, Ahmed El-Kishky, Siddharth Goyal, Mandeep Baines, Onur Celebi, Guillaume Wenzek, Vishrav Chaudhary, Naman Goyal, Tom Birch, Vitaliy Liptchinsky, Sergey Edunov, Edouard Grave, Michael Auli, and Armand Joulin. 2020. Beyond english-centric multiling… [find](https://scholar.google.com/scholar?q=Angela+Fan%2C+Shruti+Bhosale%2C+Holger+Schwenk%2C+Zhiyi+Ma%2C+Ahmed+El%2DKishky%2C+Siddharth+Goyal%2C+Mandeep+Baines%2C+Onur+Celebi%2C+Guillaume+Wenzek%2C+Vishrav+Chaudhary%2C+Naman+Goyal%2C+Tom+Birch%2C+Vitaliy+Liptchinsky%2C+S) 4. Robert Kirk Headley. 1977. Cambodian-English dictionary. Catholic University Press. [find](https://scholar.google.com/scholar?q=Robert+Kirk+Headley%2E+1977%2E+Cambodian%2DEnglish+dictionary%2E+Catholic+University+Press%2E) 5. International Phonetic Association. 2020. The international phonetic alphabet (ipa) chart. Online resource. Accessed: 2025-10-23. [find](https://scholar.google.com/scholar?q=International+Phonetic+Association%2E+2020%2E+The+international+phonetic+alphabet+%28ipa%29+chart%2E+Online+resource%2E+Accessed%3A+2025%2D10%2D23%2E) 6. Allen D Kerr. 1972. Lao-English dictionary. Catholic University Press. [find](https://scholar.google.com/scholar?q=Allen+D+Kerr%2E+1972%2E+Lao%2DEnglish+dictionary%2E+Catholic+University+Press%2E) 7. Philipp Koehn. 2024. Neural methods for aligning large-scale parallel corpora from the web for south and East Asian languages. In Proceedings of the Ninth Conference on Machine Translation, pages 1454–1466, Miami, Florida, USA. Association for Computational Linguistics. [find](https://scholar.google.com/scholar?q=Philipp+Koehn%2E+2024%2E+Neural+methods+for+aligning+large%2Dscale+parallel+corpora+from+the+web+for+south+and+East+Asian+languages%2E+In+Proceedings+of+the+Ninth+Conference+on+Machine+Translation%2C+pages+1454) 8. Jackson L. Lee, Lucas F.E. Ashby, M. Elizabeth Garza, Yeonju Lee-Sikka, Sean Miller, Alan Wong, Arya D. McCarthy, and Kyle Gorman. 2020. Massively multilingual pronunciation modeling with WikiPron. In Proceedings of the Twelfth Language Resources and Evaluation Conference, pages 4223–4228, Marseille… [find](https://scholar.google.com/scholar?q=Jackson+L%2E+Lee%2C+Lucas+F%2EE%2E+Ashby%2C+M%2E+Elizabeth+Garza%2C+Yeonju+Lee%2DSikka%2C+Sean+Miller%2C+Alan+Wong%2C+Arya+D%2E+McCarthy%2C+and+Kyle+Gorman%2E+2020%2E+Massively+multilingual+pronunciation+modeling+with+WikiPron%2E+In) 9. Nankai Lin, Yingwen Fu, Chuwei Chen, Ziyu Yang, and Shengyi Jiang. 2022. LaoPLM: Pre-trained language models for Lao. In Proceedings of the Thirteenth Language Resources and Evaluation Conference, pages 6506–6512, Marseille, France. European Language Resources Association. [find](https://scholar.google.com/scholar?q=Nankai+Lin%2C+Yingwen+Fu%2C+Chuwei+Chen%2C+Ziyu+Yang%2C+and+Shengyi+Jiang%2E+2022%2E+LaoPLM%3A+Pre%2Dtrained+language+models+for+Lao%2E+In+Proceedings+of+the+Thirteenth+Language+Resources+and+Evaluation+Conference%2C+pag) 10. Pierre Lison and Jörg Tiedemann. 2016. Open-Subtitles2016: Extracting large parallel corpora from movie and TV subtitles. In Proceedings of the Tenth International Conference on Language Resources and Evaluation (LREC’16), pages 923–929, Portorož, Slovenia. European Language Resources Association (E… [find](https://scholar.google.com/scholar?q=Pierre+Lison+and+J%C3%B6rg+Tiedemann%2E+2016%2E+Open%2DSubtitles2016%3A+Extracting+large+parallel+corpora+from+movie+and+TV+subtitles%2E+In+Proceedings+of+the+Tenth+International+Conference+on+Language+Resources+and) 11. Steven Moran and Daniel McCloy, editors. 2019. PHOIBLE 2.0. Max Planck Institute for the Science of Human History, Jena. [find](https://scholar.google.com/scholar?q=Steven+Moran+and+Daniel+McCloy%2C+editors%2E+2019%2E+PHOIBLE+2%2E0%2E+Max+Planck+Institute+for+the+Science+of+Human+History%2C+Jena%2E) 12. David R Mortensen, Siddharth Dalmia, and Patrick Littell. 2018. Epitran: Precision g2p for many languages. In Proceedings of the Eleventh International Conference on Language Resources and Evaluation (LREC 2018). [find](https://scholar.google.com/scholar?q=David+R+Mortensen%2C+Siddharth+Dalmia%2C+and+Patrick+Littell%2E+2018%2E+Epitran%3A+Precision+g2p+for+many+languages%2E+In+Proceedings+of+the+Eleventh+International+Conference+on+Language+Resources+and+Evaluation+) [our card](https://md.research.elamur.ai/free/en/papers/mortensen-2018-epitran-precision-g2p-many-languages/) 13. Maja Popović. 2017. chrf++: words helping character n-grams. In Proceedings of the second conference on machine translation, pages 612– 618. [find](https://scholar.google.com/scholar?q=Maja+Popovi%C4%87%2E+2017%2E+chrf%2B%2B%3A+words+helping+character+n%2Dgrams%2E+In+Proceedings+of+the+second+conference+on+machine+translation%2C+pages+612%E2%80%93+618%2E) [our card](https://md.research.elamur.ai/free/en/papers/popovic-2017-chrf-words-helping-character-n/) 14. Holger Schwenk, Guillaume Wenzek, Sergey Edunov, Edouard Grave, and Armand Joulin. 2020. Ccmatrix: Mining billions of high-quality parallel sentences on the web. [find](https://scholar.google.com/scholar?q=Holger+Schwenk%2C+Guillaume+Wenzek%2C+Sergey+Edunov%2C+Edouard+Grave%2C+and+Armand+Joulin%2E+2020%2E+Ccmatrix%3A+Mining+billions+of+high%2Dquality+parallel+sentences+on+the+web%2E) [our card](https://md.research.elamur.ai/free/en/papers/schwenk-2019-ccmatrix-mining-billions-high-quality/) 15. Kak Soky, Xugang Lu, Peng Shen, Hiroaki Kato, Hisashi Kawai, Chuon Vanna, and Vichet Chea. 2016. Building wfst based grapheme to phoneme conversion for khmer. Proc. Khmer Natural Language Processing (KNLP). [find](https://scholar.google.com/scholar?q=Kak+Soky%2C+Xugang+Lu%2C+Peng+Shen%2C+Hiroaki+Kato%2C+Hisashi+Kawai%2C+Chuon+Vanna%2C+and+Vichet+Chea%2E+2016%2E+Building+wfst+based+grapheme+to+phoneme+conversion+for+khmer%2E+Proc%2E+Khmer+Natural+Language+Processing+%28) 16. Jörg Tiedemann. 2012. Parallel data, tools and interfaces in OPUS. In Proceedings of the Eighth International Conference on Language Resources and Evaluation (LREC’12), pages 2214–2218, Istanbul, Turkey. European Language Resources Association (ELRA). [find](https://scholar.google.com/scholar?q=J%C3%B6rg+Tiedemann%2E+2012%2E+Parallel+data%2C+tools+and+interfaces+in+OPUS%2E+In+Proceedings+of+the+Eighth+International+Conference+on+Language+Resources+and+Evaluation+%28LREC%E2%80%9912%29%2C+pages+2214%E2%80%932218%2C+Istanbul%2C+Turk)