The Awful German Language (1880)

(faculty.georgetown.edu)

186 points nalinidash | 1 comments | 16 May 25 04:09 UTC | HN request time: 0.371s | source

Show context

rawbert ◴[16 May 25 06:21 UTC] No.44002326[source]▶

As a developer working in a German company the question of translating some domain language items into English comes up here and there. Mostly we fail because the German compound words are so f*** precise that we are unable to find short matching English translations...unfortunately our non-native devs have to learn complex words they can't barely pronounce :D

Most of the time we try to use English for technical identifiers and German for business langugage, leading to lets say "interesting" code, but it works for us.

replies(18): >>44002397 #>>44002459 #>>44002514 #>>44002534 #>>44002678 #>>44002701 #>>44002803 #>>44002985 #>>44003209 #>>44003272 #>>44003276 #>>44003429 #>>44003432 #>>44005478 #>>44005580 #>>44006867 #>>44007883 #>>44008646 #

marcosscriven ◴[16 May 25 08:22 UTC] No.44002985[source]▶

>>44002326 #

I think the issue of German compound nouns is seriously overegged. In almost all cases, it’s essentially the same as English, except with some spaces. It’s not like suddenly a short compound word expresses something that couldn’t be in English.

replies(10): >>44003194 #>>44003252 #>>44003401 #>>44003464 #>>44003598 #>>44003753 #>>44006295 #>>44006980 #>>44007613 #>>44010526 #

1. yubblegum ◴[16 May 25 14:58 UTC] No.44006295[source]▶

>>44002985 #

I wonder 'where' these compound words end up in an n-dim embedding space (relative to their German and say English 'parts'). In fact this brings up the interesting question of tokenization of the long German compound words, and how all this plays out in German to English (and reverse) LLM translation and text generation.

↑