The estimated reading time for this post is 92 seconds
We were having a discussion in the team about the various different approaches to multi language content campaigns. Should we have different accounts for each language? Should we put multi-language content in the one post?
It really depends on your brand what you’re trying to do. For Hong Kong (the market my team cared about); one account makes the most sense.
But one approach we ruled out after looking around at a selection of different retail brands was using a machine translation service in-lieu of English editorial for multi-language content.
This example came from the Facebook page of Circle K Hong Kong. Circle K is a convenience store chain that is almost ubiquitous in Hong Kong. It is a subsidiary of the Fung Group.
Despite all the discussions about deep learning there is still a way to go for mainstream multi-language content solutions. Hong Kong Cantonese is particularly hard for a few reasons:
- The language is immensely symbolic
- Cantonese uses a lot of idioms
- It has evolved very fast compared to the mainland version of Cantonese. Popular media, the internet and international culture evolved the language out of all recognition. A language expert friend listened to Cantonese language courses issued to US government personnel in Hong Kong in the early 1960s, commented how similar it was to the mainland.
Whilst Cantonese is a tough language for machine learning, it is good at seeing machine learning’s limits. English language in popular parlance has changed a lot. The use of bae, basic, finsta or lit are classic examples of uses that wouldn’t have occurred ten years ago. Google Translation got its corpus of information from looking at organisations that publish multi-language content like UN organisations. But this officialise doesn’t provide much guidance for how a language is used outside of official policy documents.