epitrochoid413

Yes I agree with you here.

LLMs tend to be a bit random, but are still more consistent and predictable than slot-machines. Also, gambling tends to exert lower effort and higher dopamine hits than vibe-coding, making it way more addictive.

But LLMs are still addictive to some extent. Maybe its around the same level as other behavioral addictions like food, social media, or gaming addictions.

That’s a fair concern. It’s a hybrid system, so dictionaries and rules handle the high-confidence cases, and the context model handles selected ambiguous words.

今日 is extremely common, but in the vast majority of ordinary text it’s read きょう. The こんにち reading is much less common and tends to appear in more formal contexts, so I handled it conservatively.

I do plan to unblock more cases like this once I find a way to speed up the context model.

Thanks, this kind of report is very useful.

如何 is context-dependent, and I hadn’t come across this case yet. I’ll add it to the model soon. Really appreciate the report and the kind words.

Thank you this is very helpful, especially from a native speaker.

今日 is a tradeoff I made intentionally: I disabled the fallback model for it because most cases are きょう, while こんにち is much rarer. But yes, this is one of the cases that gets lost with that choice.

And agreed on Japanese dictionary support. I plan to add Japanese soon. Thanks again.

Thanks, good catch. This is a known class of edge case I am trying to improve: adjacent tokens sometimes get over-merged into a phrase reading when they should stay separate. 今 and 何 should be handled separately here, not as こんなに. I appreciate this concrete example. That helps a lot.

Thanks, I really appreciate that.

No signup was a deliberate choice, to keep the barrier to trying it as low as possible.

I’ve thought about pitch accent, but it feels like a whole separate beast. The datasets are less comprehensive, and pitch can be even more context-dependent. I’d like to look into it eventually though.

Thanks for sharing this. It looks like a really cool project, and making the data public domain is especially generous.

I especially like the dictionary + example sentence format. I haven’t found a really good Japanese-English dictionary for learners, and yours looks promising.

I’m curious how token-intensive the repeated Claude polishing runs are.

Thanks, that’s great to hear. Thanks for the vouch too, I didn’t realize the comment was dead.

Thank you, that means a lot. I’ve been working on it for about a year now, so it’s really encouraging to hear it’s useful. I’m hoping to keep pushing the accuracy further, especially on the remaining hard cases like rendaku and person/place names.

I built a context-aware furigana converter for Japanese text, files, and web pages.

The main problem I wanted to solve was that simple dictionary-based furigana works well for common cases, but breaks on words where the reading depends on context:

* 市場: いちば or しじょう

* 大分: おおいた or だいぶ

* 人気: にんき or ひとけ

* 最中: さいちゅう or さなか or もなか

* 方: かた or ほう

The engine is a hybrid system:

* Sudachi for tokenization, base forms, POS, and candidate readings

* Expanded dictionary coverage for compounds and fixed expressions

* Custom rules for counters, suffixes, rendaku patterns, and phrase overrides

* ModernBERT fallback for 144 especially context-dependent target words

I have been testing it against an LLM-assisted benchmark of 7,500 Japanese lines. On the current benchmark, it gets about 12 wrong readings per 1,000 tokens. I treat that as a practical regression benchmark rather than a formal academic evaluation, but it has been useful for comparing versions and catching regressions.

The hardest remaining cases are personal names, place names, rendaku, rare vocabulary, and domain-specific terms.

I would especially appreciate examples where it gets the reading wrong, since those are the most useful for improving the system.

Same here. I use codex for planning and deepseek v4 flash for implementation. Its worked really well so far.

Meanwhile Deepseek is cutting inference costs to mere cents. Thats the real AI revolution for you.

Lets see how OpenAI holds up. They prolly shitify or dumb down their models like Anthropic to finally turn their massive loss streak into a profit.

Another round of lets dumb down the previous model so the new model feels "game changing" and "OP".

OpenAI is being squeezed from both sides.

ChatGPT's chat quality has recently dropped hard. While Claude is pricier, it actually takes the effort to think through complex tasks.

All the while, Chinese models are providing cheaper alternatives.