The problem was that we were inadvertently using the google translate auto-detect feature. Some of the bad words were being translated from their meaning in other languages to common words in English (e.g. 'hore' is auto-detected as a slovak word meaning 'up'). As a result, we had some unexpected common words in all the language lists, including English.
This change sets the source language explicitly to English, and skips translation for the English output list.
I also added a unit test for the known issues ('up' in English and 'comment' in French).
Prevent false positive results for common words (see https://github.com/LLK/scratch-vm/issues/1891).
The problem was that we were inadvertently using the google translate auto-detect feature. Some of the bad words were being translated from their meaning in other languages to common words in English (e.g. 'hore' is auto-detected as a slovak word meaning 'up'). As a result, we had some unexpected common words in all the language lists, including English.
This change sets the source language explicitly to English, and skips translation for the English output list.
I also added a unit test for the known issues ('up' in English and 'comment' in French).