context
2
ctx:discord/blah/language/2Source document
full textlanguage-2
text/plain3 KB
doc:agent/language-2/101cb3f9-915e-42b5-a9c3-9e318a3a3a81[2025-04-24 14:24] ajaxdavis: ``` Component | Practical tricks Orthography → phoneme mapping (G2P) | • If spelling is transparent, start with rule-based G2P (regex or finite-state).• Otherwise train one inside MFA or [lex4all] to auto-generate a lexicon from a handful of word/audio pairs GitHub. Pronunciation lexicon | Store in W3C PLS XML so every toolkit can read it. Text normalisation | Create small, language-specific modules for numbers, dates, kin terms, reduplication, etc. You can write them in Python and call them from ESPnet/Coqui. Prosody hints | Leverage your grammar research: encode phrase boundaries, focus particles, and tone markings as SSML pauses or <prosody> tags. ``` [2025-04-24 14:24] ajaxdavis: (files: image.png) [2025-04-24 14:25] ajaxdavis: prosody is what im looking for [2025-04-24 14:29] ajaxdavis: (files: image.png) [2025-04-24 14:40] ajaxdavis: https://gist.github.com/thomasdavis/b4786927dba78a4b52fe6f4a32b149e7 kuku yalanhi pdf to markdown worked pretty well [2025-04-24 14:40] ajaxdavis: I think the IPA a bit shit, but I think an LLM could auto correct that [2025-04-24 15:01] ajaxdavis: https://gist.github.com/thomasdavis/caf335df575ef21718926cd611c3ddd3 current plan [2025-04-24 15:07] ajaxdavis: https://gist.github.com/thomasdavis/223a994e75b3c15eb4c76230765e34d4 how to convert the markdown to linguistdic aware formats [2025-04-24 16:13] ajaxdavis: (files: nLFMhqR9.png) [2025-04-24 16:25] ajaxdavis: (files: message.txt) [2025-04-24 18:33] ajaxdavis: https://john.mccr.ae/ [2025-04-24 18:36] ajaxdavis: (files: image.png) [2025-04-25 01:33] lisamegawatts: Pama–Nyung [2025-04-25 01:35] lisamegawatts: https://pmc.ncbi.nlm.nih.gov/articles/PMC2981908/ [2025-04-25 01:36] ajaxdavis: epitran [2025-04-25 01:40] ajaxdavis: https://github.com/australia/mobtranslate.com/blob/master/dictionaries/kuku_yalanji/resources/grammar.pdf [2025-04-25 01:40] ajaxdavis: this is the grammar file that i've turned to markdown, and then ran through o3-mini in chunks to create different file formats [2025-04-25 01:40] ajaxdavis: https://github.com/australia/mobtranslate.com/tree/master/experiments/pdftomd/extract/output [2025-04-25 01:41] ajaxdavis: the xitg file contains the IPA data [2025-04-25 01:42] lisamegawatts: https://github.com/xigt/xigt [2025-04-25 01:43] ajaxdavis: https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner [2025-04-25 01:45] ajaxdavis: ng -> nnnnn [2025-04-25 01:45] ajaxdavis: kuku -> gugu [2025-04-25 01:51] ajaxdavis: https://generous.rocks/chat/f1423ad6-e2f5-4009-9e80-58b2e98e1d64 [2025-04-25 01:54] ajaxdavis: https://www.isc.ro/ ^ this is where the elites play [2025-04-25 02:16] optionalsecurity: (files: image0.png) [2025-04-25 02:16] optionalsecurity: Somehow I got a generous generous when I clicked the PLAY NOW! button [2025-05-07 00:15] ajaxdavis: https://news.ycombinator.com/item?id=43905299
Facts in this context
Grouped by subject. Each subject links to its full article.
Ajaxdavis23 factsex:ajaxdavis
| postedMessage | Message 2025 04 24 14 25 |
| postedMessage | Message 2025 04 25 01 54 |
| postedMessage | Message 2025 04 25 01 51 |
| postedMessage | Message 2025 04 24 14 24 |
| postedMessage | Message 2025 04 25 01 45 2 |
| postedMessage | Message 2025 04 25 01 45 |
| postedMessage | Message 2025 04 25 01 43 |
| postedMessage | Message 2025 04 25 01 41 |
| postedMessage | Message 2025 04 25 01 40 3 |
| postedMessage | Message 2025 04 25 01 40 2 |
| postedMessage | Message 2025 04 25 01 40 |