creating a dictionary in my native language

[itswck]

Apologies in advance for my ignorance in this area. I'll get straight to my point:

I would like to investigate how I could contribute to this project and add the characters for my native language. I believe that 'Dictionary creation' is what I need? I have had a look at the section on dictionary creation in the repo and (as someone with no experience here) I'm struggling to understand what exactly I need to do and why.

Could somebody please explain (or send a resource) what is actually required in order to add characters from a language that is currently not included in the app? I can see some steps to follow here but, again, I'm not able to conceptualise exactly what's required of me and why.

Thank you.

[Helium314]

What you need is a wordlist like this example. You can find some more information in the dictionaries repo.

Once you have the list, you use the AOSP dicitonary tools to compile a dictionary from the wordlist, like java -jar dicttool_aosp.jar makedict -s <language>_wordlist.combined -d main_<language>.dict.

If you know how to use python, you can try the (still a little experimental) scripts for creating a wordlist from a list of sentences or words.

[netizeni]

Hi @Helium314,

Thank you for maintaining this great piece of FOSS software! I have one question related to the topic.

There's a dictionary of a language I would like to use in the dictionaries' repo, but it's in Cyrillic script, while I would like it to use in Latin script. As I never did something related to this, what do you suggest being the best/fastest approach to contribute and solve this, i.e. taking the existing dictionary in one script and transforming it in another?

[netizeni]

So you want to transliterate the cyrillic script to latin?

Pretty much that. I just wanted to see if there's any way to feed some software with wordlist.combined file (or even better .dict file) without the need to sanitize it first i.e. to remove word=,f=218,flags=,originalFreq=218, and it does transliteration automatically.

[Helium314]

Probably the easiest way is writing a short python script. that for each line

1. extracts the word
2. calls the transliterate function
3. replaces the original word with the tranlsiterated one

[netizeni]

@Helium314 I've managed to create a wordlist with all transliterated words, compiled the dictionary with java jar, but when I try adding it to the desired language, it shows "Error: script not compatible with this keyboard".

I used an already existing wordlist, tried both changing and leaving the description, date and version unchanged in the header, but I get the same error. Do you maybe know what's causing the problem? The language is Serbian (Latin) and the wordlist is attached below:

wordlist.zip

Another thing I noticed, which is more or less related to this. For Slavic languages (maybe some others I didn't check) and QWERTY layout, if the lowercase letter is pressed and hold, in a popup it shows a variation of that letter in the local form, for example if "s" is pressed it shows "š" or "c" shows "č". Meanwhile, when the letter is in uppercase it shows only a symbol, i.e. letter "s" shows #.

I wanted to confirm is this a bug, or it requires some setting to be changed?

[Helium314]

The problem with Serbian is that the default script is cyrillic. The app uses sr_zz for Serbian with latin script, if you change the language code accordingly it should work.
(This might change soon. After dropping Android 4 support, we can switch to language tags like sr-Latn to explicitly specify the script)

I wanted to confirm is this a bug, or it requires some setting to be changed?

That's a bug. It's already fixed, but a version with the fix is not yet released.

[netizeni]

That fixed it, thanks! I will open a PR soon to add it to dict repo.