Automatic Transliterations

Great feature, thank you for implementing it! As Kurdê Dîn said, some letters appear weirdly: I saw Я appearing as YA instead of Ya. There may be others, but if I see them, I'll write another comment.

As for suggesting other languages, Bulgarian would definitely need to be added to the list. However, there is not only one system of transliterating it, so maybe adding two or possibly three ways would be best. Should I write a comment here or send you a PM explaining in detail my suggestions?

There is no standard for transliteration of foreign languages in using Chinese characters. It could be "sound" right but no meaning at all.

For my experience of learning foreign languages, IPA is my priority. And the Romanization is good enough for me to pronounce the foreign language. If I'm using Chinese characters to pronounce the foreign language (actually I tried to read them in Russian lyrics), it just makes me more confused. So I like to read the transliteration of Romanization.

Question from the user: if the automatic transliterations credit to the lyrics submitters, do they get the point too?

I have to say that today, on my 7th anniversary here, reading this makes me happy :)

It's amazing to see how much LT grew up and got new functions with time. It's a pleasure to be a part of this community!

On the way to our millionth translation!

We're getting old, pal, but let's not talk about our ages here :P 😂

After observing a while the romanisation systems for Bulgarian, I've noticed some issues. I'll separate them accordingly:

First system (closer to the official one):

1. Хх currently appears as Kh/kh – it would be way better to appear simply as Hh because it has no k in it as sound.
2. I noticed Цц romanised as Ts/ts, whereas тс – as t·s, with a separating dot in the middle to distinguish it from the other one. I understand why it's done like this, but the difference, in my opinion, is barely noticeable (unless bolded: t·s/t·s) and therefore misleading. Moreover, although similar, Цц is a single sound, whereas тс consists of two separate sounds, thus they're not the same, so I'd suggest Цц to be Cc, (like it's done in other Slavic languages using the Latin script, e.g. Serbian, Czech, Slovak), whereas тс to become ts without a separating dot.
3. Ъъ is currently Ŭŭ, but I think it will be better to appear as Ǎǎ, because as a sound, ъ is closer to а than to у.
4. Юю and Яя appear as Yu/yu and Ya/ya respectively: nothing wrong with that. Йй, however, appears as Ĭĭ, which isn't that wrong, but Yy would be way better because the letters ю and я represent й+у and й+а respectively, and in this way they will be consistently transliterated.

Second system (more phonetic one, closer to other Slavic languages using the Latin script):

1. Ъъ currently appears as ", which is very, very unusual. I've never seen it transliterated like this anywhere, plus it doesn't give a clue how one should pronounce it, so given the more phonetic nature of this romanisation, the best way for it would be Ǎǎ, just like in the other system.
2. Шш appears as Šš – nothing wrong with that. Щщ, however, appears as Ŝŝ, which I've never seen like this ever before, plus I doubt it anyone would notice the difference between the caron and the circumflex, which, being a small one, would be very misleading, as in Bulgarian, Щщ is ш+т, so the best way for it is to get transliterated as Št/št.
3. Юю and Яя appearing as Ûû and Ââ respectively: it's not that wrong, but it's pretty unusual, to say the least. Given that Йй is Jj in this system, Юю and Яя should be rather Ju/ju and Ja/ja respectively.

And lastly, I noticed both systems lack a transliterated ѝ (и with a grave accent) – it should get transliterated as ì in both systems (You may add, just in case, the uppercase Ѝ (Ì) as well, but it's never actually used).

EDIT: I also just saw ь transliterated as an apostrophe, which is also strange for Bulgarian. Since it represents the same sound as й does, ь should be transliterated as y in the first system and j in the second one.

EDIT 2: A few more vowels, just in case, would be better to have their transliterations too: А̀а̀ (Àà), О̀о̀ (Òò), У̀у̀ (Ùù), Ѐѐ (Èè), Ю̀ю̀ (1: Yù/yù; 2: Jù/jù), Я̀я̀ (1: Yà/yà; 2: Jà/jà).

I can see your point and I'm aware that the same would apply for other Romance languages, like Italian or Portuguese for instance. At the same time, however, not many languages lack the H sound and its allophones; in fact, Romance languages are well-known for lacking it (with the exception of Spanish [Jj] and Romanian [Hh]), but compared to many other languages from different groups, they appear to be the minority and shouldn't therefore have much of an impact on transliterating languages that feature this sound. All Slavic languages share it, English, German, Arabic, Finnish, Hungarian, Mandarin, Japanese, etc, do so too.

French (and probably other languages) may transliterate lacking sounds in different ways, but that doesn't mean Bulgarian specifically does it the same way. Kh is kind of outdated and it would be strange for a native speaker to type it like this. It's also seen from a very Romance-based point of view, which doesn't have much to do with representing a Slavic language's phonetic system, as I already said. Plus, a transliteration should also look as natural as possible to a native speaker – despite the fact it's usually meant to serve as a guide for non-natives.

Apart from that, if the first system is meant to be closer to the official one for Bulgarian (the current one is from 2006), then Хх should be Hh (Hristo Botev, Harmanli). An ideal romanisation system is difficult to achieve, but I'm rather prone to think that whoever is interested in learning how to pronounce this or that letter/sound in a certain language, can always look it up on Google or Wikipedia where IPA guides with audio examples are to be found.

Thank you very much! I just checked them and everything's perfect, with just a few small things you seem to have missed, they're all in the first system:

• Capital Х appears as Kh instead of H
• ѝ doesn't appear as ì, but just ѝ
• Йй still appears as Ĭĭ instead of Yy
• тс appears as t·s instead of ts

Meanwhile, I was wondering if the letter Жж (in the first system again) should get transliterated as Jj rather than Zh/zh.

The combination zh can be understood as зх, which is a common combination of letters in frequently used words like разходка (razhodka), разхубавил (razhubavil), разхайтен (razhayten). If put in the opposite direction, zh can appear ambiguous and therefore words like the ones I gave could be understood as ражодка, ражубавил and ражайтен instead, making no sense at all. Another example would be the word изживявам, which currently will appear as izzhivyavam, not making it clear for people who can't read the Cyrillic script whether the zzh should be pronounced as зж, ззх or even жж.

If you decide so, you can leave it the way it is currently, but in my opinion, Jj would make this ambiguity disappear; plus, the majority of Bulgarians do use Jj for Жж when typing in Latin script, so it's not uncommon practice at all.

Yes, these are Bulgarian vowels with grave accents (for comparison, Russian uses acutes, e.g. и́). They're used in rare occasions (with the exception of ѝ, being a word itself) to mark where the stress falls in words you can't tell for sure how they should be pronounced, for distinguishing words whose written forms are the same, but their pronunciation differs (e.g. напра̀ви (He/she did) – направѝ (Do!), or when a word is not pronounced the standard way (common practice in poetry and lyrics). That's why it would be best for these vowels to be transliterated as well, just in case.

(Sorry for my extreme perfectionism and detailed explanations, I'm just trying to make things clear and as less ambiguous as possible.)

Excellent, just checked it and everything's flawless. Thank you very much again.

Another example:

https://lyricstranslate.com/en/%E3%81%95%E3%81%8F%E3%82%89%E8%B2%9D%E3%8...

Sari ikeru kimi ni sasagen [Sariyukeru] (could be both 'i' and 'yu', no way to tell but to listen)
Kono kai wa kyonen no hamabe ni [kyonen --> kozo] (no way to tell, out-dated usage)
Ware ichi nin hiroi shi kai yo [ichi nin --> hitori] (this one is usually hitori)

Honobono to usu beni shimu ru wa [shimu ru --> somuru] (classical Japanese)
Waga yuru samishi chishio yo [yuru --> moyuru] (classical Japanese)
Wa roba ro to kayou kaori wa [wa roba --> harobaro]
Kimi koi uru mune no sa zanami [koi uru --> kouru] (classical Japanese) [sazanami - one word, could be fixed]

I think it would be helpful, as long as it's possible, to add names to the automatic translations. Calling them "Romanisation 1" and "Romanisation 2" doesn't really help. I'd suggest to put the name of transliteration system in question. It gets more technical, but who's specifically looking for transliterations would find it useful.

Different transliteration systems have different names. As you can see from other comments, transliteration is mainly based on choices: should a letter be transliterated following their pronunciation? Or would it be better to choose one unambiguous letter for each letter or the other language's alphabet? Some choices are also made for historical reasons.
For istance, here you can see a table of different systems of transliteration for Russian, the way they differ and their names.
This is an example for Greek.
As I said, a downside is that this would get too technical, but at least everything would be categorised and well ordered.

I'm not sure, it might be better to have something for songs that don't have a transliteration for people who just want to get an idea.
Maybe add a note for the Japanese romanization that there may be minor mistakes or something like that? (or in the faq, idk)

It could be a start to eventually having one that works much better.
You could also collect somewhere the mistakes that are found, and fix them when you're able.
Or when some ML does the job, there'll be some things to test it on ;.)

I'm not familiar with other non-alphabetic scripts beside the Chinese language(s), and for Chinese google translate does an excellent job in transliteration.
It's just the history of the Japanese writing system that makes it so.. difficult to automate.
So I don't agree with the blanket statement.

The rest I wrote in the previous comment.
On the other hand, it does bug me that it's all wrong

It would be nice to have Automatic Translation and then a function for users to be able to edit the auto, because it gets most of it and would save a lot of time.

I'm really agreeing with this sentiment, for Japanese.

Even though there is a disclaimer that there might be errors, it feels more like there *will* be errors. Japanese songs can get very creative and there are many ways to read or write different kanji and words to express what the singer wants whether through context or just having a better flow to the song. So they can end up using unusual, unconventional or even archaic words/phrases. Automatic Romanization will never be able to catch these little quirks. If a user looks up the song and can't even sing along properly without a human made transliteration what's the point of leaving up automatic Romanization?

It is also detrimental to amateurs as they might use this as a crutch, not realizing that the automatic Romanization is wrong and they base their English translation on that. But... they shouldn't be translating anyways if they don't catch such errors. But it's always possible this could happen.

I feel it's better to remove Japanese automatic Romanization and keep relying on humans for transliteration.

@lt, thank you for implementing it!
I checked a few lyrics and noticed the following issues:

• Latin J j should be Ј ј in Cyrillic, not Й й;
• Lj lj and Nj nj should be Љ љ and Њ њ respectively;
• Dž dž should be Џ џ, not Дж дж.