update readme for integerized LSTM models

[Shreeshrii]

Will upload the traineddata files next.

[amitdo]

GitHub upload failed for files > 25mb

Which files?

[Shreeshrii]

I have created another PR with 2nd set of fles, please merge that too. I will list the languages that could not be uploaded after that. ShreeDevi

…

____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com

On Thu, Mar 22, 2018 at 3:50 PM, Amit D. ***@***.***> wrote: GitHub upload failed for files > 25mb Which files? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#88 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AE2_o8M5TkESg8wXKD_Ncvo_rs-Fo6sAks5tg3r-gaJpZM4S2uav> .

[Shreeshrii]

So far, the files that did not upload are: bod chi_sim chi_tra jpn kan khm lao mya san The older version of files are still there. ShreeDevi

…

____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com On Thu, Mar 22, 2018 at 4:29 PM, ShreeDevi Kumar <shreeshrii@gmail.com> wrote:

I have created another PR with 2nd set of fles, please merge that too. I will list the languages that could not be uploaded after that. ShreeDevi ____________________________________________________________ भजन - कीर्तन - आरती @ http://bhajans.ramparivar.com On Thu, Mar 22, 2018 at 3:50 PM, Amit D. ***@***.***> wrote: > GitHub upload failed for files > 25mb > > Which files? > > — > You are receiving this because you authored the thread. > Reply to this email directly, view it on GitHub > <#88 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AE2_o8M5TkESg8wXKD_Ncvo_rs-Fo6sAks5tg3r-gaJpZM4S2uav> > . >

[stweil]

@Shreeshrii, how did you handle deu_frak.traineddata and the other files which have no best traineddata?

[Shreeshrii]

@stweil I did not have a check whether file exists in best. Just the version string would have gotten updated.

[Shreeshrii]

@stweil

Is it just the three user contributed files , dan_frak, deu_frak and slk_frak?
Is there a way to go back to an older commit only for these files (functionality is still the same, just the version string is changed) or should I reupload an older version?

[stweil]

No, I don't think that this is necessary. We can keep them as they are.

[amitdo]

kur has no lstm. Does it have Latin or Arabic letters?
kur_ara is from best.

[Shreeshrii]

I looked at langdata repo just now. It has both kur and kur_ara. Looks like there was a change in langcode but the files were not moved.

I had taken the list of RTL languages from language_specific.sh, it did not have kur, but has kur_ara.

langdata/kur has training text, wordlists etc, in Arabic script. While langdata/kur_ara only has a list of desired and forbidden characters. Hence the kur_ara traineddata file in tessdata_best is not correct. Probably same will apply to tessdata_fast - I haven't checked.

I will file an issue in langdata mentioning this. Hopefully all this will be fixed when Ray/Jeff update langdata for 4.0.0.

[Shreeshrii]

kur has no lstm. Does it have Latin or Arabic letters?
kur_ara is from best.

@amitdo Thanks for bringing notice to this.

https://en.wikipedia.org/wiki/Kurdish_languages says

In use:

Hawar alphabet (Latin script; used mostly in Turkey and Syria)

Sorani alphabet(Perso-Arabic script; used mostly in Iraq and Iran)

Not used:

Cyrillic alphabet (former Soviet Union)

So probably both kur and kur_ara can be there with appropriate langdata.

[amitdo]

An older issue about kur: #45