In a previous article about track titles, I dissected the structure of a classical music track title, from the composer’s name to the movement title. I closed it article on an open question about translations:
Another implication is that titles are translatable. A Symphony is a common noun in English and could be translated as Symphonie in French or Sinfonia in Spanish. The same applies to « Choral » and « in D minor ». In French, this track could be said as « Beethoven : Symphonie n°9 en ré mineur Op. 125 – Chorale – 1. Allegro ma non troppo… », identifying the exact same work.
In this article, I am scratching the surface of title auto-translation. In most cases, translating automatically leads to hazardous results. However, given the formal structure of these titles, I am giving a try. Why?
- In order to parse and language process existing content, using the target language will naturally lead to better result.
A comparison analysis
In this post, I will use the following example:
This is the formal English title of the 13th Shostakovich symphony. It has a usual name “Babi Yar”, called after the Yevtushenko poem used in the first movement. I highlighted the different parts that are stored as fields in my Rondo DB database.
The French way to display this title would be:
- The number field (13) and catalog field (Op. 113) are not translated. They generally won’t be (as long as we remain in the same alphabet) as they are related to number, people names (catalogs such as K., BWV. or BB.) or in this case the worldwide used Latin word: Opus (means: work).
- Babi Yar it not translated as well, because it refers to a work that is not translated. However, we can expect that this field can be translated, and cannot be auto-translated.
- Symphonie and si bémol mineur are translated fields. However, from a common noun dictionary, we could legitimately expect a semi-auto-translated effort. All Symphony in English would fall into Symphonie in French, resulting in a one-shot translation effort.
- Chostakovitch is also translated, although a person’s last name. It is frequent that Russian names are written very differently from a country to another, because Cyrillic to Latin alphabet is not commonly shared from one language to another. Non-Cyrillic names can also have various translations, because first names are translatable (Johann Sebastian Bach vs Jean-Sébastien Bach) but also depends on local practice. This obviously can’t be auto-translated, but as a consolation one translation is shared throughout one composer’s catalogue.
- We observe that the order of the fields remain the same from English to French
- We also observe that in English No. is used to shorten the word Number, while n° is used in French.
Let’s have a look at translations of this title in Italian and German:
- I’ll be honest, I like the Shostakovich case 🙂 Look at all these beautiful translations! (I got them from Wikipedia, it’s very hard to know when you’re not a local classical music fan)
- In this experiment, German is the only language inverting the number and the work type fields. The dot after 13 must be read as th.
- Their is another subtlety in German. When the mode is minor (Moll) the preceding key is written lowercase. When the mode is major (Dur) the preceding key is uppercase (like in G-Dur).
Looking at this example can lead to a positive answer to an auto or semi-auto translation effort. However music titles are more complex than that and it’s hardly possible to automate all titles translation.
Nevertheless, I would like to emphasize, based on this example, some key points:
- The structure of the title is different from one language to another, but formal and supposedly identical to all works of the same type (Symphony).
- Some fields don’t translate, typically numbers (in this alphabet).
- Some fields are semi-auto-translatable, with a high factorizing ratio. Rondo DB, for instance, counts 211 piece types, 17 keys and 12 modes. Translation effort is a one-shot for these fields.
- Some fields need a specific, unique translation. Composer’s last name (though reusable for all pieces in its corpus) and piece name (Baba Yar being a counter-example).
It’s more complicated
Symphony is a simple case. I am currently under further investigation, and I can say that some problems arrive quickly. Some examples.
- Concertos makes sense only when attached to an instrument. Concerto No. 1 could be for piano or for violin as the solo instrument, in one composer’s given catalogue. They will both have No. 1 in common, and will be distinguishable by an instrument indication (Piano Concerto No. 1 vs Violin Concerto No. 1). This leads to more syntax and translation effort (but seems still possible).
- Some titles are not compound titles, typically opera titles. They are more like a film title, and exactly like film titles they can be translated or not depending on the country.
- There are exceptions. I am thinking of Sinfonia, from Luciano Berio. It’s a Symphony, and Symphony is translated as Sinfonia in Italian ; but this piece is still called Sinfonia in other country (at least in France).
If you think of other funny examples, let me know 😉