Translations
I translated 10 of my favourite songs by Asian Kung-Fu Generation and tried to live out my dashboard dreams
I had planned on translating the entire playlist of 23 songs, but I kept adding and removing songs from the playlist. At around song ten I noticed that the rules I’d set myself were the least fun part of the whole thing, so I dropped them. I still translate songs; they just don’t end up here. So consider this sunset. Everything below is the complete record of those ten.
- Google Sheets: where I logged every minute and left myself notes.
- Spotify API: song names and track lengths, straight from the playlist.
- kanji-data: JLPT level, frequency and stroke count for every kanji.
- Flask and plain JavaScript. No charting library; the charts are hand-drawn SVG (by Claude, not by hand, I’m afraid).
Asian Kung-Fu Generation is a Japanese alt-rock band from Yokohama and my favourite band. I spent around 4 months translating their songs when I had time. I got to practise language and get some measurements. The workflow consisted of:
- Writing down the lyrics out by hand, highlighting the words I don’t know, and breaking them apart. Example.
- Run a script to pull out features: track length, unique character count, the JLPT level of every kanji, minutes spent, etc.
- Pray to God I wasn’t wasting my precious time on this earth
Minutes per unique character, in the order I translated them. Lower is faster. It mostly goes down, which is a relief. Hover/Tap a point for the song.
Minutes spent on each song.
Unique characters against minutes spent.
Minutes per unique character for each song. Shorter is faster.
Japanese has 3 different alphabets because, somehow, they never had a literacy crisis and had to simplify like the rest of the world. One of them is Kanji, which are split into difficulty levels N5 → N1. I wanted to know how much of my time went into each level.
A kanji that turns up five times in the chorus counts five times. JLPT hasn’t published official kanji lists since 2010, so the levels come from the community lists everyone uses anyway. And recognising a kanji is not the same as reading the word it’s in, as I was reminded roughly once a line.