Suggestion for Japanese lingq

for what it’s worth, I agree with amop567 - I think it will take you longer this way.

I don’t agree with the analogy regarding the ‘b’ in the word subtle being made above. You can still perfectly understand someone who mispronounces the word subtle (in some dialects of English around the world for all I know, this may even be correct pronounciation). In Japanese you are simply learning a word that does not exist. You can understand the meaning of a character (which in principle can be read without knowing any Japanese), yes, but you don’t know the japanese word (which I think is the most important thing, since it’s the language you’re learning). It takes a bit more effort to go and look up the correct pronounciation, but I think it pays off in the long run as you are learning both the meaning of the character and the actual Japanese word.

There are also so many mistakes in the furigana that I also think it would be a real task to start unlearning all of that. To me a more correct analogy would be like a Japanese person reading English, deciding he can’t figure out the pronounciation of the word “subtle”, and therefore calling it “popo” (or some other word that doesn’t exist) instead; just so he can give it some reading in his head. If that would be the only word, no problem, but if he is doing that for 10-20% of his vocab, I’m convinced that will be a real challenge to unlearn.

That being said, I also started at an intermediate level and not from scratch- if that is how it works for you, who am I to say otherwise :slight_smile:

1 Like

@azarya Thanks for clarifying. Mistaken furigana (like replacing a kunyomi word with the onyomi, for example) will leave you with a gibberish word that does not exist in Japanese.

I am surprised to learn that people find it helpful, but of course, to each their own. As I said above, ftornay’s idea of a warning message might be a better way to keep the feature available for those who like using it without leading other users astray.

As it stands, a beginning learner with little knowledge of the Japanese writing system would assume that the hiragana feature provided at least reasonably accurate readings when this is not actually the case.

I get the feeling this could alienate many prospective users which would be a real shame.

Additionally, there is a certain kind of logic in a lot of on-yomi readings as the majority of Kanji are partly phonetic ideographs (see for example: https://www.kanjiclinic.com/kc5final.htm#:~:text=Kanji%20containing%20phonetic%20components%20are,that%20signals%20the%20on%20pronunciation)- meaning there is a hint on how it is pronounced in one part of the kanji. While not the case for all Kanji, the more you are exposed to the correct phonetic reading of kanji, the easier you can guess what similar kanji sound like - by not applying correct reading, I would imagine you slow down this process.

@azarya
Thank you for your advice. I was aware that many kanji have a phonetic component to them, based on old Chinese pronunciation. It’s good to know that paying attention to it can in fact help recognize kanji even for Japanese (onyomi) pronunciation.
Just to clarify: I don’t need to “unlearn” anything. I check the pronunciation for every new word I encounter. I use the hiragana to help me read words I already know in both pronunciation and meaning but whose kanji I haven’t learned yet.

Thank you for your reply and your comment. At the very least it will be a reminder for us beginners that we should mind this issue from the beginning. If you do not mind me asking: did you find wrong furigana totally wrong for that kanji? I thought it marked a kanji reading out of context, but it was a valid reading in a different context.

As a compromise, in the reading options, where the “furigana” can be selected, a message could be introduced warning that it is not guaranteed that the furigana is correct (in that context ? / in any context?)

For example:
→ “人たち(ひとたち) becomes じんたち”
At least じん is a valid pronounciation in other cases, right?

Are there examples in which the furigana is completely off, no matter the context?

The furigana readings you get are always possible pronunciations of the kanji. The issue is that the algorithm doesn’t always select the reading that fits the context. It is related to difficulties in word parsing.
Your example is in fact typical, “jin” is one of the onyomi pronunciations of 人

Thanks for the feedback. I leave furigana off and have since the beginning. Yes, sometimes it’s inaccurate and that will be improved for later versions of LingQ.

Personally, I think furigana hinders progress (romaji to an even more so). However, there are beginners who want furigana, which is why the option is there. It may not be 100 percent accurate but LingQ encourages you to LingQ so you can find various meanings and how to say the word. Many times, the Japanese definitions in LingQ show the word in kana form.

Thanks. That is what I thought. So they are not totally wrong and useless, they are wrong in that context. I think that I will manage to learn the different pronounciations in their context with time.

It’s wrong in that it leaves you with a nonsense word that is unintelligible in Japanese. To me, that is wrong in every practical sense of the word.

1 Like

Thank you Eric. Looking forward to seeing the improvements,

@Jokojoko83
Since you’re interested in using the automatic furigana, let me explain what IMO the issues are.
Bad readings tend to occur when kanji are isolated. That is, everytime Lingq considers there is a word that consists of one single kanji with no other kanjis and no okurigana. In that case, the system simply guesses the reading. It’ll always give you an existing pronunciation, but it’ll often be the wrong one for the current context. For some reason, the algorithm seems to favor onyomi readings in those cases, which, IMO, is not the best default option.
Such isolated kanjis can, in turn, occur in two cases:
a) Because the word is in fact written with one single kanji. This is the least frequent case. For example this happens with numbers (which are often mis-transcribed) and some help-words: prepositions, conjunctions, …
b) Because the parsing fails and a kanji gets isolated from the rest of the word. This is by far the most usual issue. In such cases not only is the furigana incorrect but the translation as well and you have to create a lingq that links together the disjoint parts, whether or not you’re using automated furigana.

Thanks Ftornay, that was a clear explanation

Have any steps been taken to fix the problem? It seems that the only way to correct the incorrect furigana would be to do it by hand, which shouldn’t be very hard, considering that all one needs to do is to compare the text to the recording. Even a relative beginner could do the work. Why not let a group of users who are studying the texts anyway start making corrections? I would be glad to volunteer to start fixing the mini stories…

Brilliant answer. Indeed, I have found myself crossing out furigana in paper books when I actually know the kanji but don’t trust my own knowledge enough not to glance at the furigana. So I will now also switch off the furigana in Lingq, especially as it’s sometimes just plain wrong anyway.

In defense of LingQ, it should also be said that many words simply have different pronunciations. Even native Japanese speakers of use more than one pronunciation for no specific reason. I see that all automated tools make similar mistakes.
That said, I am with Eric. I have never turned on the hurigana and romaji. They are largely an impediment, not a help. I do use a form of romaji when I create a LingQ for myself, because it is easier/less time-consuming to type but that is as far as I go.

You make a good point. The first English speech synthesisers (now usually called Text-to-Speech engines) made plenty of prounciation mistakes. English is pretty much like kanji with letters: the letters give an idea of what the pronunciation could possibly be like, but it remains not much more than a wild guess.