More free and cheaper translations!

[vBET]

Hello.

At this moment we are making research about new translations APIs. The goal is - more free and cheaper translations for you

Right now you can use fee translations with vBET. This is given by Apertium and Microsoft APIs. Still Apertium do not cover lot of supported language pairs, and Microsoft is free with limitations (for more you need to pay).

Anyway vBET allows to use all supported engines at once, to minimize your translations costs (so use Apertium if possible, then use Microsoft if still have free limit, then...), so we want to add new APIs to make it even cheaper.

We are already experimenting with SDL BeGlobal which is still in sandbox, but soon will be fully open and are cheaper than Google. Also we found some other FREE APIs (with limitations, but vBET will switch to another engine automatically, when limits will be reached).

So assuming right now we started some tests with:
- Yandex API (FREE translations with high quotas: 1 000 000 characters/day! To compare Microsoft gives free 2M characters/months)
- DetectLanguage API (FREE language detection with high limits, user can choose paid higher limits)
- SDL BeGlobal API (paid translations, but cheaper than Google)
- Gengo API (translations, not sure is it paid or free, for sure human translation is paid, se send them email with questions about machine translation - in examples is 0 cost, in API result there is no info... - but we are watching close to this one)

We also still have some other APIs to check. There is lot of others. Unfortunately many of those allows only one translation per one request, what is not acceptable in our usage, where we translate lot of small portions of text stripped out of HTML, but if you know any translation/language detection API just write here. We will check it for possibility to use in vBET

[vBET]

And we already made beta integration with Yendex API. Right now it is running on this forum, so we will be able to check error logs if any appears and made correction before release
Tomorrow we will first test it on vB3.x test forum and then put it on live forum with vB3.x

[vBET]

We just added new API for language detection named: DetectLanguage. It is great option, since it gives 5 000 free requests/day and recognizes all supported by vBET languages. It allows to buy higher limits, but vBET uses detection only for new messages, so as long as your forum do not have more than 5 000 new messages daily you can have language detection completely free.

At this moment on this very forum we are already testing DetectLanguage, also Yandex and SDL BeGlobal translations APIs. So next release will be great improvement for all our users
Those are also tested on real vB3.x forum.

Before we release we want to monitor it for a while and maybe discover some other APIs which are able to use with vBET

[vBET]

We also made logs investigation and detected that messages blocked by anti-spam adds was also used in language detection. For some forums this can be lot of traffic. Since new release vBET will use language detection just before writing new message into database and only if other plugins didn't tell that it should be abandon. We used way already bound in system so no special integration will be needed at all. For example we tested it with plugin No Links or Images Allowed for People with Fewer...

[vBET]

We seriously investigate right now where translations come from and we found another update which will limit your expenses. We found out that each banned user produces new translation on his profile site. Since for many forums spammers are really big issue and there can be dozens of banned users daily, we made another update. In new release vBET will not show user pages for banned users, just error message about not registered user. First it will make better performance (what for show banned users profiles, where they can still show their info about ICQ and such things) and will not produce new translations (in page description is changed text then which is different for each user so it produces new translation each time).

This will be configurable and turned on by default.

[tavenger5]

Thanks for the updates!

[tavenger5]

any update on when 4.7 will be released?

[vBET]

We were planning on the end of this month. Still it is with assumption that SDL BeGlobal will update their API in next days and allow for table translation. They planned this for middle of this month, and still nothing - so we will see.

Anyway at this moment w already have added also other APIs - Yandex and DetectLanguage. Booth free with big quotas. Also some tests are made with Let's MT, which is free, but seems to be not ready for big traffic, so we are not sure will it be included in next release, later or there will be some bigger problems.

Still one issue on our side, after some translation optimizations in area of not translating numbers, some bugs appear in RSS channels and now not translated area is translated with broken markups. So this is one must do on your side.

Please note that for several weeks we already test it on this live forum and log every new translation which is send to provider. This way we found translations produced by changing numbers, and removed those by wrapping numbers by for example <i> markup. So on translation page such text with number in template/phrase is separated so it goes to translation in parts, and numbers are not translated (when you set the option). Also the option is not working correctly in older versions which is already corrected.

So there will be really big improvement and translations costs will be cut drastically. Still we do not hurry with it, because we made lot of changes on low integration level and we do not want to produce other meaningless translations (like now we have in RSS in our not released version).

So please be patience. We also want to release it, because it will give hope for lot of our users who limited translations because of costs.

[vBET]

And the RSS issue is already corrected. At this moment we are waiting only for SDL BeGlobal array translations. Beside this, we are ready for production

[vBET]

Today I send question to SDL BeGlobal, when arrays translation will be supported. It should be done already and we really want to include SDL BeGlobal in next release (1/4 less price than Google for translations). We will let you know when we get answer. Beside this we are ready to release.