Raphaël Merx

Machine Translation (MT) Technology in Health: A Scoping Review

Where does MT show promise in health contexts? Where is it too risky? And are its uses well understood? (spoiler alert: not really)

Started a PhD @ University of Melbourne

Machine translation; low-resource; health education; and LLMs. Supervised by Ekaterina Vylomova, Nick Thieberger and Hanna Suominen

Instituto Nacional de Linguística Seminar, 2024

Presentation “Oportunidade no dezafiu tradusaun automatika iha lian Tetun” (“Opportunities and challenges for machine translation in Tetun”) at the annual seminar for the Timor-Leste National Institute of Linguistics.

Low-Resource Machine Translation through Retrieval-Augmented LLM Prompting: A Study on the Mambai Language

Can LLMs (like ChatGPT) learn to translate into Mambai, if given relevant examples? Answer: somewhat if examples are very close to the sentence to translate, otherwise not really. Paper accepted by EURALI.

Asia-Pacific Medical Education Conference 2024

Preliminary findings from a scoping review on machine translation for health, as part of a symposium "AI, Medicine and Medical Education in Resource-Constrained Settings", organised by Dr Lois Hong, Maluk Timor clinical director.

tetun.org: more languages, more users

Added Indonesian, Portuguese, and even Tok Pisin (in beta) to tetun.org! Also passed 60,000 monthly active users.

Q&A on DFAT reports, powered by AI

Think chatGPT, but backed by hundreds of reports, and citing its sources. Reports from the DFAT (Australian Dept of Foreign Affairs) website

A 3-min pitch on why jargon in aid is worse than we think

Went to beautiful Canberra for the 2022 Australasian Aid Conf, and made a small pitch about why jargon is worse than an annoyance, and why overusing it makes us no better than AIs.

Talk about machine translation for Tok Pisin at the LSPNG conference

Presented my work on machine translation, with a first translator prototype for Tok Pisin at https://tok-pisin.org, at the Linguistic Society of PNG 2022 conference.

Talk DFAT to me

An AI trained to write in the style of Australian Aid reports.

Regional indonesian language translator

A translator between Indonesian and several regional languages of Indonesia: Javanese, Sundanese, Balinese, Acehnese, Banjarese, Buginese, and the Minangkabau language.

Tetun Wordle

A Wordle all in Tetun, forked from the Indonesian Wordle, Katla. Been told it's the first word game using Tetun? Thanks to Adérito Correia (INL) for reviewing my Tetun instructions.

Balinese spell checker

A webapp to fix Balinese text spelling, built in partnership with the Balinese Wikisource (WikiPustaka) community. Useful to fix text transliterated from the Balinese alphabet to the Latin alphabet.

Agile, meet adaptive: How global development assistance and software engineering both thrive on an iterative approach

I believe in iteration (and perseverance) in both software engineering and development assistance. Written for the Catalpa blog.

PyconAU talk on the tetun.org machine translator

Mel Mistica and I presented on how we setup a volunteer-led English-Tetun machine translation service. Since then we passed 25,000 monthly active users.

Beyond Compliance: Approaching government transparency as an Opportunity rather than an Obligation

Finding alignment of incentives between decision makers, data officers and citizens can make or break the success of transparency projects.

Added machine translation to tetun.org

What started as a Tetun-English dictionary now includes a translator, using a machine learning model for translation.

Published an iOS version of tetun.org

Over the first few months of the pandemic, I taught myself React Native and migrated the tetun.org app to it. This let me generate an iOS build which was published to the Apple App Store.