992647 - Submitting wordlists for word prediction for Irish, Scottish Gaelic and Manx

Status: RESOLVED FIXED

(Firefox OS Graveyard :: Gaia::Keyboard, defect)

Description

[Michael Bauer]

We have wordlists for all three locales and would like to include them for MOS to improve text entry for all three.
Struggling to find documentation of format, maximum size and what way is used to weigh/rank the suggestions.

Comment 1

[Jan Jongboom [:janjongboom] (inactive)]

See the _worldlist.xml files in gaia for information on the format. Maximum size is not a big issue I think, try to keep the en-US wordlist as a reference. I don't know what the weighing algorithm is though.

Comment 2

[Michael Bauer]

Thanks I'll check that out. 

Weighting - there must be some sort of weighting/ranking so for example if I start typing "typ" I want the suggestions for "type" and "typing" to be shown in preference of "typology" or "typification" which are much less common as words.

Comment 3

[Michael Bauer]

Attachment: ga.js

Based on the discussion in bug 1007547 I gather we need to submit keyboards for ga, gd and gv to begin with.
I've done three .js files based on the en.js in the repo, Kevin said they look fine to him so I'm attaching them. Please let me know if I've missed/messed up something.

Comment 4

[Michael Bauer]

Attachment: gd.js

Comment 5

[Michael Bauer]

Attachment: gv.js

Comment 6

[Michael Bauer]

Attachment: xml files for ga gd and gv

I tried following the instructions jan gave but I got myself into a muddle. I created some fork (quite some time ago) trying to run what he said:

1. Go in the gaia directory to apps/keyboard/js/imes/latin/dictionaries
2. Open Makefile
3. Append the name of your dictionary to the list (e.g. if it's called sv_wordlist.xml), put:
 ru.dict \
 sk.dict \
 sl.dict \
 sv.dict \
 sr.dict \

4. Make sure the _worldlist.xml file is in the folder
5. Run 'make sv.dict'

But best as I can tell, it didn't work as I can't see them either in my inadvertant fork (https://github.com/akerbeltz/gaia/tree/master/apps/keyboard/js/imes/latin/dictionaries) nor the main project (https://github.com/mozilla-b2g/gaia/tree/master/apps/keyboard/js/imes/latin/dictionaries).

I'm attaching the xml files, perhaps it would be easier for everyone if someone who knows what they're doing did this?

Comment 7

[Jan Jongboom [:janjongboom] (inactive)]

Taking it.

Comment 8

[Jan Jongboom [:janjongboom] (inactive)]

Attachment: Patch

You can test it out by attaching a phone to your PC and running (in the gaia directory):

GAIA_KEYBOARD_LAYOUTS=en,ga,gd,gv APP=keyboard make install-gaia

And restarting the phone. Enable the layouts in Settings->Keyboard

I tested the ga & gd, but don't know anyone who speaks gv :p

Comment 9

[Michael Bauer]

Hey that's great, thanks.

I'm stuck though, the way I read it I need to connect my Flame, cd onto it and in /gaia run that entire command above.

I've never cd'd onto a phone on Ubuntu... normally USB devices just pop up but the Flame doesn't. If I do $ sudo lsusb, it shows a Qualcomm device but no drive name. Spent half an hour googling this problem but can't figure it, could you give me a hint? :)

Comment 10

[Jan Jongboom [:janjongboom] (inactive)]

Hi Michael, check this out. https://runawaydev.wordpress.com/2013/04/22/how-to-connect-your-firefox-os-phone-to-a-linux-computer-for-app-pushing/comment-page-1/

Comment 11

[Michael Bauer]

Hi Jan!
I'm not getting anywhere, even though I spent the entire afternoon trying to get it to run. I installed all the bits those pages said I needed and followed all the steps but I must be missing something because the system refuses to let me into the device, it tells me it's there when I lsusb but that's as far as I go. Even my adb error message for ./adb devices doesn't match, I just get a totally black line.

Will this appear in the nightlies? If it does, maybe I should just wait a couple of days, put a nightly beta on and test it that way?

Or for a rough test, I could send you a sentence for each language with words that are likely to be tricky if something isn't working?

I just don't understand enough about this stuff at the developer level, I'm a linguistic expert, sorry.

Comment 12

[Jan Jongboom [:janjongboom] (inactive)]

No, this will not appear in the nightlies by default. If you can ping me on irc.mozilla.org (channel b2g or gaia) during the CEST day I'm happy to assist.

Comment 13

[Michael Bauer]

Thanks for your help Jan! The good news is that on the language side, it all works beautifully.

There's a few keyboard glitches though:
1) I altered the currency ranking in line with each locale, so the gd.js has
'£': '€ $ ¥',
which ought to put £ first followed by the others on long press? But I'm only offered $, nothing on long press.
ga.js should be (I think the £ in first place was my mistake but the long press again shows no options)
'€': '£ $ ¥',
gv.js should be
'£': '€ $ ¥',

I mentioned in the chat that all three use hyphens and apostrophes more like word characters than punctuation. I'm thinking that without messing around with the keypad the most elegant solution would be to change the . longpress from
'.': ',?!;:'
to
'.': '-',?!;:'
for ga and gd and for ga to:
'.': '-,?!;:'

But I'm not sure if having the ' in there breaks something?

Comment 14

[Rudy Lu [:rudyl] (inactive after 2015/7/3)]

Comment on attachment 8468440 [details] [review]
Patch

Looks good to me but have some nits to be addressed.
Please help take a look at the comments on the pull request.

Besides, I think we need to add these layouts to the mapping file,
build/config/keyboard-layouts.json

Thank you.

Comment 15

[Michael Bauer]

Rudy, I'm afraid I have no idea what that means (about the pull request). And I don't know how to add layouts to the mapping file. In any case I wanted to ask first if it's possible to add ' to the . longpress menu.

Comment 16

[Jan Jongboom [:janjongboom] (inactive)]

Fixed nits and comments from Comment 13. Landed in https://github.com/comoyo/gaia/commit/4d6a16f7670f2a6cd8ca651520415b575e42dce2