Romanization walkthrough

Romanization Process

For almost all cases, the romanization process is as follows:

• Obtain correct reading;
• Apply VocaDB’s romanization rules and convert the reading into romaji (English alphabet).

Step 1 can be extremely difficult at times. This guide will focus on Step 1.

Websites such as 初音ミク Wiki regularly supply the correct reading to a vast number of songs. In cases which the reading is supplied, the editor’s effort is greatly simplified, having to perform only the task of converting kana into romaji. It is highly recommended to always check the reading with 初音ミク Wiki or another trusted source. See Section 3 Part II for details.

Step 2 is a simple process of following the guideline rules.

Basic Examples

A large number of songs have very easy to read titles. These are just as simple as turning kana into romaji.

• ホシノウタ -> Hoshi no Uta

• まとい -> Matoi

Song titles which are direct loan words are also easy to romanize (recall that VocaDB uses the original language for loanwords):

• パラレルワールド -> Parallel World

Titles with common kanji are romanized as one would expect:

• タイヨウの子 -> Taiyou no Ko

• 月と風船 -> Tsuki to Fuusen

It is expected that every editor who edits romaji be able to easily output the same romaji for the above song titles. For readers who had trouble, it is recommended they do not attempt to provide their own romanizations on VocaDB.

Part I - Kanji

Kanji is an integral part of the language. It is important to remember that individual kanji and kanji compounds can have multiple readings, but in most cases, there is only one correct (*) reading (and thus only one correct romanization).

(*) Correct in this context means intended (by the artist). To illustrate, reading 静寂 as “Shijima” is technically correct, and maybe correct in rare cases (it is a reading found in a dictionary), but unintended for most cases. “Seijaku” is far more common and more likley than not what the artist intended. It is safe to assume that a more common reading is the correct reading, because the artist would have specified otherwise.

Examples:

• 蒼のクジラと電波塔 -> Ao no Kujira to Denpatou

Note that 蒼 is “Ao”, and not any other reading, such as “Sou”. Kanji reading takes a lot of experience, but often there are clues that lead to the correct reading. For example, kanji by itself takes the kun reading in general. In contrast, compound kanji words, such as 電波塔 are generally read by their on readings.

• 赤心性：カマトト荒療治 -> Sekishinsei: Kamatoto Araryouji

荒療治 is an example of an exception to the on reading rule, where 荒 is read by its kun reading. Many times it is impossible to tell what the correct reading is, without prior exposure to the term.

As seen above, it is easy to make a mistake by applying general rules and common kanji readings. Luckily, dictionaries include many compound terms. Therefore it is imperative that editors look up words they have never encountered before, even if the individual kanji is known.

Part II - Rendaku

Rendaku must be carefully considered for many Japanese words, especially nouns. For readers unfamiliar with the concept, the Wikipedia pages linked below may be a good starting point.

• 連濁 (Japanese)
• Rendaku (English)

Simply put, when two words are combined, the first character of the latter part of the combined word can either be voiced or unvoiced (adding or not adding dakuten). Rendaku rules are complex and at times unpredictable. Knowing the general rules, as well as having a strong vocabulary and high listening and speaking level is helpful in determining the rendaku outcome of a compound word. While Kanji do not indicate rendaku, they are marked in kana with dakuten. Therefore, rendaku rules for kana titles do not pose any difficulty to the editor.

*からふる日和 -> Colorful Biyori

If the title is first separated into からふる and 日和, one may incorrectly romanize the title as “Colorful Hiyori”. However, due to rendaku, 日和 is pronounced びより and is thus romanized as “Biyori”.

• ながれ星 -> Nagareboshi Rendaku is voiced on 星. Since 流れ星 itself is a word, there should be no difficulty for this romanization.

• 夜間飛行 -> Yakan Hikou Rendaku does not apply here. 飛行 is read normally as “Hikou” (and not “Bikou”).

• 夢見心地 -> Yumemi Gokochi Rendaku is voiced on 心地, changing “Kokochi” into “Gokochi”. Since 夢見心地 is a single term, “Yumemigokochi” is also correct.

Sometimes the rendaku effect is ambiguous for certain compound words. In such cases, the editor may choose the more common reading. Unless the artist has stated a particular reading which confirms the rendaku, we assume that both readings (unvoiced and voiced) are correct. Note that this is only applicable to ambiguous readings. The above examples have no ambiguity, so departing from the example produces an incorrect romanization.

Part III - Play on words and multiple meanings

Irregular readings from the creative use of words, such as puns and portmanteaus can be difficult to notice. A strong vocabulary is required to identify non-standard readings.

• 恋距離遠愛 -> Renkyori En’ai This title is a play on words of 遠距離恋愛, swapping the 遠 and 恋, producing a double-meaning.

• 音偽バナシ -> Otogi Banashi This is a play on words of おとぎ話. An editor unfamiliar with the term may incorrectly romanize the title, since “oto” is a kun reading, and “gi” is a on reading.

• 右に曲ガール -> “Migi ni Maga-ru” or “Migi ni Ma Girl” The ガール in the title is both the okurigana to 曲 (がる->ガール) and “Girl”. Since it is shared, neither interpretation is more correct than the other (which is the whole purpose of these kinds of wordplay).

Part IV - Uncommon and Forced readings

When an artist specifies a reading for a title, it must be used for the romaji reading. Editors should make a habit of quickly scanning the media description to see if any reading was specified. If a reading is supplied and it contradicts with the common or conventional reading, the editor should leave an edit note to facilitate easier reviewing.

• 好き！雪！本気マジック -> Suki! Yuki! Maji Magic

In the artist’s blog, the reading is clearly stated.（読み：すきゆきまじまじっく), so the conventional reading of 本気, “Honki” is incorrect in this context.

• 心海 -> Kokoro Umi

At first glance, an editor may be inclined to state that the reading is “Shinkai”, a wordplay of 深海. The PV of the song also takes place under the sea further suggesting that “Shinkai” is correct. However in the description of the artist’s instrumental self-cover video, a URL to a file “kokoroumi.zip” is linked. In this case it is safe to assume that the file name corresponds to the reading of the title.

Part V - Miscellaneous

Part V lists other miscellaneous rules an editor should be aware of. These exceptions maybe be uncommon but do occur.

The below three titles contain the same exception arising from historical kana usage.

• マヨヒドリ -> Mayoidori Note that ヒ is read as “i”. In kanji, it would be 迷い鳥.

• 回ル想ヒ -> Mawaru Omoi

The same concept applies here.

• あなたのおもひで -> Anata no Omoide

This is similar to the above examples. In kanji, this would be あなたの思いで.

These are easy to figure out knowing the words 迷い, 想い, and 思いで.

Part I - Edit Notes

Edit notes should be included when an editor:

• Takes a particular stance on an ambiguous reading
• Uses a reading that is uncommon or not obvious at first glance. E.g.
  • Reading found in the PV
  • Reading found in the video description
  • Reading obtained from sources such as the artist’s Twitter and official blog entries.

Leaving an edit note allows a reviewer to quickly check the romanization correctness without having to repeat the efforts of the previous editor. If there will be other edit notes (unrelated to romaji entry), the editor can choose split a large edit into smaller chunks to make it easier for others to review the edit history. For complex cases requring a detailed explanation, leaving a comment in the discussion section of the song is encouraged.

Part II - Methods, Dictionaries, Tools and Resources

Questions: