slightly offtopic: I would be delighted if Tucholsky had written something hilarious along those lines. But he didn't as far as I know, he focused more on the German language himself.
I'm a Korean currently learning Japanese, and while I do understand that this mash of Chinese characters and kana writing system can be appreciated for its exoticness, as a learner I can't help but feel it's more of a hassle resulting from it being a not yet fully optimized writing system. (I mean, do we really need both hiragana and katakana?)
I'm definitely not claiming that Korean is a more "optimized" language overall, but at least when it comes to the writing system, we had exactly the same problem as the Japanese (if you look at Korean newspapers just a few decades ago they are littered with Chinese characters), and at some point we fully ditched Chinese characters and have no problem going on with our lives. In fact, it made our lives easier in many cases, especially in keyboard typing.
As a side note, we obviously have some side effects from switching to entirely phonetic alphabet system. For example, the words "tea" and "car" have the same pronunciation (차=cha) and so they are indistinguishable in writing, but it wasn't the case when Chinese characters were used (茶/車). I'm not sure how this side effect propagates into some sort of sociolinguistic phenomenon, but at least for average people it doesn't seem to have much significance.
>I mean, do we really need both hiragana and katakana?
It is pretty redundant, but it's helpful sometimes: foreign loanwords (esp. from English these days) are almost exclusively written in katakana, so they stand out. Also, having some stuff in katakana and some in hiragana helps to distinguish the different words, since Japanese doesn't use spaces.
>For example, the words "tea" and "car" have the same pronunciation (차=cha) and so they are indistinguishable in writing, but it wasn't the case when Chinese characters were used (茶/車)
1. I'm surprised "car" isn't called "sha" instead (as it is in the onyomi reading of 車 in Japanese, which was borrowed from Chinese).
2. This would be a good place to borrow from Japanese: the kinyomi (stand-alone word) pronunciation is "kuruma", which probably won't be confused with anything in Korean. :-)
As for Korean ditching Chinese characters, it seems like Japanese is doing that too, just very very slowly. There are many words that have kanji versions that no one uses any more, preferring hiragana instead. And there's a lot of stuff being borrowed from English. For instance, 切符 (kippu) is the normal Japanese word for "ticket", but these days, everyone's calling it "チケット" (chiketto) which is borrowed from English.
Watching Dr. Eric Lander's lectures in MIT OCW Biology course was a watershed moment in my life. I felt exactly the same way as this blog post after finishing the course, two years after graduating from college. Highly recommended if you haven't seen it yet: https://youtu.be/P-Ry4rRdDbk?si=Hp2SEkEKGeeui-pP
May I ask what library you're using for kanji to hiragana transliteration? I'm working on a language product as well and I'm using pykakasi which is certainly prone to errors. I tried your app and noticed similar errors as well (大いばり should show ooibari instead of daiibari, for example). Wonder if we can do better on transliteration.
MeCab with ipadic and a lot of custom swift logic for fixing issue patterns and matching to JMDict entries as an additional heuristic that the stemming/token was done right. I’m also using a custom generated JLPT classification (a more complete guess at what the full set of JLPT vocab is based on ebook word freq) to choose more likely candidate results. I haven’t improved this in a couple years, it’s one of my upcoming focuses now that I have the app rewritten and out.
Unidic also interesting but harder to use and huge data size.
I’m going to be layering on gpt to further improve.
Given enough advances in hardware and software optimization, isn't it reasonable to think that if we connect this level of language model to speech-to-text + image-to-text models on the input side and robotic control system on the output side, and set up an online end-to-end reinforcement learning system, the product will be a convincingly sentient robot, at least on the surface? Or am I underestimating the difficulty of connecting these different models? Would like to hear from the experts on this.
Hey! we did 2 events in Taipei in 2022 w/ 30-40 ish people attending each time. We will try to keep it running regularly in 2023. Happy to join force if you want to help making it happen more regularly!
This is a very short-sighted view. Globalization requires certain level of operational efficiencies, such as speed and cost of communication and transportation. We have just recently reached a level where governments and corporations can support demands from billions of population. Technology will keep on progressing unless we end up in an extinction scenario, and at certain level globalization will be just another step in humanity's perpetual drive towards centralization.
It's actually a very long term view, looking at the last 9000 years or so of history we are aware of. You also have requirements confused with incentives. That the things you mention are required does not in any way mean globalization will happen if they are present. There needs to be a strong incentive that is more valuable than the status quo to all the parties involved and there are all sorts of assumptions about that incentive that are unclear if they are true or pretty clear they are false (eg: assuming economies of scale will continue with size at a meaningful enough value rather than plateauing, changing slope to be much flatter or outright decreasing. Assuming technology will always improve. Assuming population changes from, for example, an ultra low birth rate won't precipitate one of the many breaking up of large countries we have seen in history, etc.)
