Proxy Translation and Common Localization Blunders

What is a proxy when it comes to translation? It’s like putting a layer over your site, where site content is scraped and then translated on the fly through a third-party proxy server. Generally seen as simple and quick to get up and running, proxy solutions don’t require much IT to get started. However, I recently came across a site that implemented a Japanese translation proxy solution quite poorly, so I thought I’d demonstrate with screenshots a couple things to consider, not just for proxy solutions, but also for common localization issues in general:

1. Image text will not be picked up by the proxy. This screenshot shows that text within an image file won’t get scraped by the proxy crawler for translation, and depending on how the site is set up, certain text even not in images can be missed, too! So, there’s a fair amount of risk for “bleedthrough” of the source language, and this might not be the best thing for a potential customer to see.

2. Even with certain tactics to make image text findable by the proxy crawler, formatting issues still remain. This picture shows what happened when the proxy crawler WAS able to pick up the text in a poster (due to the text being overlaid on the poster, not actually part of the image file), but then failed to put the translated text back in properly. You can see the Japanese is cut off, and there wasn’t any consideration put into the font either. And also, note that with a lot of marketing material, you’d likely adapt the actual imagery/design of posters too rather than just the text, so a proxy would limit the ability to make more major changes between regions.

3. Text expansion issues – obviously not something confined to just proxy solutions, but worth noting anyway as a general localization issue. In this side-by-side, you can see the Japanese text expanded to fill three lines, causing it to be cut off.

4. Machine translation – this screenshot shows a laughable translation of “sales” into Japanese, where the resulting word 売上高 is the kind of “sales” you’d only see in a financial report. This was obviously machine translated, and although MT issues aren’t proxy-specific, it’s worth mentioning since many proxies do use MT. The price listed for a proxy solution may just be for MT only, and the resulting mistakes from that can make it so the site is worse off than if it hadn’t been translated at all. Other related points to consider: does the proxy solution segment text in a way that will make for a high quality TM? (does it HAVE functionality for TMs to begin with?) Can you sync the proxy’s TM to another platform that you use for translation of other files for the company?

Anyway, lots to think about. Love always being able to learn more about all the different ways content is localized and analyze things when I see them pop up around me!