A workflow for amateur translators

[feihong]

I recently published the 50th post on the Chinese Lyrics Challenge blog, so I thought I might be qualified to write a brief guide on how to do a large amount of translations if you're doing it as a hobbie rather than a profession. I'm not going to go into a ton of detail, because if I try to write everything down, it'll take forever and I'll never finish. If you're curious about something, feel free to ask questions in the comments.

 

As a bit of background, the Lyrics Challenge blog is only concerned with translating the lyrics from Chinese songs to idiomatic English. It's not a playlist of the songs I like the most, rather, it's meant to be a list of songs whose lyrics I find interesting. I prefer songs that have a storyline, evocative imagery, or both. Most of the songs on the blog are specific enough that they cannot be confused with any other song. The checklist below is specific to lyrics translation, but the overall process should apply to any kind of translation.

 

1. The first step is to pick the right songs (lyrics) to translate. I don't care if a song has the best melody, but it better not be too grating because I'll be listening to it many times. I really don't like lyrics with a lot of abstract wording. This causes the space of possible translations to explode and just fries my brain. The best choice is a song whose general meaning you understand, but there are several lines that you don't quite understand. I discover new songs to translate mostly by checking my YouTube feed and following the hot songs playlist on StreetVoice.
2. Get the lyrics. Usually this is pretty easy, just do a search in Google. However, often you have to convert from traditional to simplified. There's actually no really good service I found to do this, but I use this one because it's less bad than others: https://www.purpleculture.net/traditional-simplified-converter/ (but I still have to manually replace 著 with 着).
3. Proofread the lyrics. It's common for lyrics you get from the internet to contain typos and very often they will contain grammar mistakes. I try to fix as many of these as I can before I start the actual translation process.
4. I convert the lyrics document to a table in Notion (a fancy wiki site). The table columns are Chinese, English, and Annotation. Each line gets its own row. I actually tried a lot of sites besides Notion, but it won out in the end because it's easy to share with collaborators and the mobile app works well enough (it has a lot of problems, but other apps were worse). It's a lot of manual work to do the conversion from plain text to table, so I wrote a Python script to automate the process.
5. Now the translation actually begins. I'll occasionally translate an entire song in one go, but typically it happens in spurts. Most of the translations are written through the Notion app on my phone. Sometimes I'll only do a single stanza or even a single line before putting it down to attend to some other business. A lot of times, I'll force myself to write something, anything, even if I know it's wrong. There will be plenty of opportunities to fix mistakes during the review process. For me, the translation process is a marathon, not a sprint. Better to spend 10 minutes on it each day rather than 70 minutes once a week. This is especially true when a translation requires a lot of dictionary and thesaurus lookups, which fry my brain rather quickly.
6. After the initial draft is done, it's time to review. However, I prefer to let some time pass before starting review. This gives your subconscious mind some time to mull over the lines you wrote. It's better to start another translation rather than start a review immediately after finishing the first draft. Once review starts, I'll go back and carefully examine each line. Sometimes my mistakes will be obvious, other times I'm not sure so I add some notes in the Annotation column.
7. If you want to produce translations of a decent quality, you must have editors. I schedule a weekly call with some friends who grew up in China and I read out each line in Chinese while they check the line's English translation on Notion. They suggest edits on the spot and one of the nice things about Notion is that when I change the document, it updates in real-time for everyone viewing the document. Some additional grammar issues in the original lines are fixed at this stage.
8. Now it's time for another review! Yes, even though the editors already went over it, it's possible that they only pointed out problems but weren't able to come up with good solutions. Usually it's a problem with phrasing. Something that is easy to express in Chinese can be devilishly hard to communicate in non-mangled English.
9. The last step is to publish the translation. I also semi-automate the process with a different script, this time converting the Notion table to HTML that can be posted in the blog editor on this site. One of the reasons the script saves so much time is that it allows me to only write one translated line for a line that repeats multiple times. I only go through the publishing process once every few months, because I do a lot of them at once and then schedule them to be published at the rate of once a week for the next few months.

 

I think that's a good summary, so I'll stop here. This describes the general workflow that I follow now. Occasionally the process is a little different, for example if the lyrics can't be found anywhere and I have to transcribe them myself. If you have any suggestions for how I could improve the workflow I've described here, I'd love to hear them.

 

P.S. There are 50 posts published so far, but there are currently 7 more scheduled to be published in the coming months. And about 50 more translations that are in various stages of the translation pipeline. My goal is do about 100 translations, which will take about 2 years in total. If I maintain the current rate of publishing, I should reach that goal by August 2021.

[roddy]

Something you don't mention, although I suspect it happens in the proofreading stage, is just having a read of the document before you start. While on rush jobs I might simply have to start translating before reading the thing, it's obviously far better to read it first. Ideally, with decent chunks of time between reading, translating, and reviewing. 

[feihong]

1 hour ago, roddy said:

Something you don't mention, although I suspect it happens in the proofreading stage, is just having a read of the document before you start.

Yes, the first complete reading occurs at proofreading stage of the original document. But this reading is cursory, I don’t bother to look up any words or phrases at this stage. I don’t do a deep reading until translation starts.

[roddy]

I might be inclined to do those look-ups a while before you start translating. Make sure you've got a full understanding, let it ferment in your brain for a while, then do the translation as a purer translation task, rather than dictionary work and so on. 'Course, that's in an ideal world. 

[feihong]

24 minutes ago, roddy said:

 

I might be inclined to do those look-ups a while before you start translating. Make sure you've got a full understanding, let it ferment in your brain for a while, then do the translation as a purer translation task, rather than dictionary work and so on.

 

I agree this is far more efficient, but I rarely do this. I’ve gotten bogged down at times because many songs require a lot of lookups and I start procrastinating because it feels like I’m not making progress. So my process puts more priority on maintaining momentum rather than efficiency. For example, if I get stuck on a line I’ll just skip it. If I feel like I’m stuck on a song I’ll switch to another song. It takes longer overall but at least I keep going.
 

But there are rare cases where the lyrics are a bit too much over my head. In these cases, I’ll go over the problematic lines during the call with my editors. In this scenario I haven’t yet started translating, I simply take notes in the Annotation column of the table.

[roddy]

18 hours ago, feihong said:

I agree this is far more efficient, but I rarely do this.

I have that in needlepoint over my desk.