opensubtitles-uploader icon indicating copy to clipboard operation
opensubtitles-uploader copied to clipboard

Published 20 hours ago • verified publisher icon vankasteelj

Many Issues language detection

Open matant94 opened this issue 2 years ago • 2 comments

In Advanced Subtitle Station format: The app fail to detect the Spanish (Spain) language Fails to recognize the differece between Portuguese and Portuguese BR (same as 3 types of chienese) Sometimes recognize English as Icelandic

matant94 avatar Mar 17 '22 18:03 matant94

Could you provide the files that are misidentified? It's not AI-powered so it has to rely on manual tweaks and predictions using a dictionnary, it's not perfect (nor will it ever be), but it might be tweaked better

vankasteelj avatar May 10 '22 21:05 vankasteelj

Not detected (English).zip Detected as German instead of English.zip

tbertels avatar Jun 20 '22 22:06 tbertels

@tbertels your examples are fixed in the sources

vankasteelj avatar May 10 '23 21:05 vankasteelj

@hajimekun without your files I cannot do anything more.

vankasteelj avatar May 10 '23 21:05 vankasteelj

I've just had a new "Files.detectSubLang() error: not conclusive enough (anonymous) @ files.js:112" error (with the new version), here's the subtitle: Not detected (English) 2.zip

tbertels avatar May 11 '23 14:05 tbertels

The probability of "english" was below the 25% treshold. For that specific file, it's 24.16% detected as english. FYI: Only 72 sentences were detected as "english" in your file, but the algorithm also thinks 40 sentences are in "scots". There is nothing more I can do at that point. I can't go lower on the treshold or it will cause issue and risk misinterpreting subtitles lang. The point is : be sure or don't auto-fill. Not "guess hoping it's right"

vankasteelj avatar May 11 '23 19:05 vankasteelj