Notebook

A Zettelkasten kept in the open, in four languages. Translations and poetry in Breton, essays in Welsh, and working notes on language technology — corpora, machine translation, speech, lexicography — most of them written while the work was still going wrong.

Recently All notes (106) →

1. 7 Sept 2026 br Administrer et étendre Microsoft 365 Copilot
2. 6 Sept 2026 br 6 a viz Gwengolo 2026
3. 6 Sept 2026 br Deizlevr
4. 6 Sept 2026 fr La corruption
5. 3 Sept 2026 br Des réseaux de neurones aux LLMs
6. 3 Sept 2026 br 2022-2026, panorama de l'écosystème IA, couches technologiques et principaux acteurs

• Deizlevr — er yezh a teu din
• Blog — in whatever language comes to me

Brezhoneg

• Troidigezhioù
  • Tonkad ar Yezh — Saunders Lewis
  • Notennoù diwar-benn an droidigezh
  • Ar Bisklavret — Marie de France
  • Suzanne — Leonard Cohen
• Barzhoniezh
  • Trid an tridi
  • Kreac’h
• Yezh ha politikerezh
  • Penaos saveteiñ ar brezhoneg
  • Opiom an intellektualed
  • Mojenn war-giz
  • Dialektik ar renadoù
  • Teiryezhegezh evit kargidi Europa
• Yezhoniezh
  • Pellgomz pe bellgomzer
  • Gerioù keltiek er yezhoù slavek
  • Notennoù war Meurgorf

Cymraeg

• Ysgrifau
  • Ieithoedd Lleiafrif yn Oes Data Science a Dysgu Peirianyddol
  • Anodd yw creu geiriau Cymraeg
  • Deallwydd Sgyrsiol
• Tystysgrif Sgiliau Iaith
  • Sesiwn paratoi 1
  • Sesiwn paratoi 2
  • Sesiwn paratoi 3
  • Sesiwn paratoi 4

Teknologiezh ar yezh

• Corpora and data
  • A Breton–French parallel corpus from scanned novels
  • Coqui TTS + XTTS
  • Language technology portfolio
• Models and methods
  • The evolution of transformers
  • Ar matrisoù evel rouedad
  • Kelenn an AI
  • Hayes & Wilson on maximum-entropy phonotactics
• Evaluation
  • Evaluating a customer support agent
  • Sentiment analysis
• Liammoù
  • LinkedIn posts
  • Leksis — the omnilingual dictionary
  • Data is better together

Formation

• Programmes
  • L’IA au service des développeurs
  • RAG et fine-tuning
  • M365 Copilot
• Libre
  • La corruption

Prederiañ

• Testing and psychometrics
  • Why we need yet another testing framework
  • Psychometrics for dummies
  • Elo rating and the Rasch model
  • Reviewing Pelánek 2016
  • Notes on optimization
• Teaching and learning
  • How should we teach languages
  • Code-switching and language level
• Atomoù
  • Zettelkasten
  • Aesthetic is the elegant solution of a problem
  • Conflicting constraints
  • Long distance dependencies as direct relationships

Ober

• This site
  • Publishing an Obsidian Zettelkasten with Astro
  • Every note, by language and year
• C’hoariellyoù
  • Jankó Piano — an isomorphic piano you can play on your computer keyboard
• Cool links
  • Draw neural nets

Diellaoueg — M.Sc. Technolegiaeth Iaith, Bangor 2024-25

• Dissertation
  • Dashboard
  • Research proposal
  • Experimental design
  • Dissertation ideas
  • Acknowledgements
• ICE-4005 mini-project
  • Dashboard
  • Report
  • Repository
  • Last meeting log
• Research methodology
  • Review of an outstanding dissertation
  • Reviewing Lemhöfer and Broersma (LexTALE)
• NLP labs
  • Lab 1
  • Lab 2
  • Lab 3
  • Lab 4
  • Corpus report
  • Final exam
• Technoleg Iaith
  • 0. Cyflwyniad
  • 1. Dysgu Peirianyddol a Chyfieithu Peirianyddol
  • 2. Testun i Leferydd
  • 3. Adnabod Lleferydd
  • 4. Deallwydd Artiffisial Sgyrsiol