Raphaël Merx @rapha.dev · 18/10/2025Whoa the #WMT25 results on MT Evaluation are wild! ChrF outperforms pretty much all neural metrics 🙀 100
Raphaël Merx @rapha.dev · 27/07/2025in Vienna for ACL, presenting Tulun, a system for low-resource in-domain translation, using LLMs Tuesday @ 4pm Working w 2 real use cases: medical translation into Tetun 🇹🇱 & disaster relief speech translation in Bislama 🇻🇺 131
Raphaël Merx @rapha.dev · 25/05/2025(3) The vast majority of usage is on mobile (over 90% of users / over 80k devices) Takeaway: publishing MT model in mobile apps is probably more impactful than setting up a website / HuggingFace space. 120
Raphaël Merx @rapha.dev · 25/05/2025(2) Translation into Tetun is in higher demand (by >2x) than translation from Tetun Takeaway for us MT folks: focus on translation into low-res langs, harder but more impactful 100
Raphaël Merx @rapha.dev · 25/05/2025We find that (1) a LOT of usage is for educational purposes (>50% of translated text) --> contrasts sharply with Tetun corpora (e.g. MADLAD), dominated by news & religion. Takeaway: don't evaluate MT on overrepresented domains (e.g. religion)! You risk misrepresenting end-user exp. 100
Raphaël Merx @rapha.dev · 25/05/2025Our paper on who uses tetun.org, and what for, got published at the LoResMT 2025 workshop! An emotional paper for me, going back to the project that got me into a machine learning PhD in the first place. 230
Raphaël Merx @rapha.dev · 08/05/2025My favourite ICLR paper so far. Methodology, findings and their implications are all very cool. In particular Fig. 2 + this discussion point: 041
Raphaël Merx @rapha.dev · 05/12/2024Our paper on generating bilingual example sentences with LLMs got best paper award @ ALTA in Canberra! arxiv.org/abs/2410.03182 We work with French / Indonesian / Tetun, find that annotators don't agree about what's a "good example", but that LLMs can align with a specific annotator. 010
