Universal Translator

Source code to the core parts of my Universal Translator posted to the LSL Wiki.

https://wiki.secondlife.com/wiki/Universal_Translator

Anyone want help with the upgrades so that we can make a better world-wide translation system?

My hero. Your translator is a fine piece of work, and you are a prince for making the code available.

Thanks for sharing that with the community Hank.

My brainstorm is that Translation Services should be built into the Server Side. it would work pretty much the way your Translator does now, but without the overhead of being LSL executing on the mono engine.

Anyway, what features do you think you will be adding next?

They're working on putting translation into the client in Snowglobe.

Bizarrely, the translator in Snowglobe skips the one area where it could have offered a real improvement over LSL translators, the ability to reach into IMs.

This trend to reinvent scripted wheels as viewer features keeps turning up a common theme: the C++ version is invariably less flexible or configurable, and it's definitely less convenient to recompile a whole viewer to add a favorite feature than it is to save a tweaked script. A couple of spots where Snowglobe translation is kind of inflexible are in deciding if only certain objects or avatars will be translated, and how to decide what the source language is.

They could add scripting to the viewer. It already has a Javascript interpreter embedded.

Not a problem! Some features I'd like to see, and are outlined on the wiki, are...

1. Support for Google Spellcheck. Would be great to be able to spell-check chat first, and automatically do suggested corrections before the translation.

2. Break the engine script into two or more scripts to allow for future expansion and to be able to handle low memory situations better.

3. Support pseudo-instant message. This could be done with avatars wearing the device within 100m of each others, or could be done world-wide using email or some other communications mechanism. Avatars would chat on a hidden channel, and would be able to communicate with another avatar or group chat. Would require inworld servers of prims to link translators up together and find one another.

4. Ensure that unicode is supported correctly. Ensure that Right-to-Left languages are input and output correctly. May need additional options since the SL client doesn't support this correctly and sometimes people speak backwards into SL chat to get around this.

5. Rework auto-detection of language code. Should only automatically change the avatar's language if it is reliably predicted that they have changed languages.

6. Add in automatic "spam" detection and prevent translation. Find way to detect gesture "chat spam" and filter that out from the translation. Filter out bad translations that only return the same or similar input. Support replacement of common abbreviations or slang so that the google translation system can translate correctly. May require support of a user-definable dictionary notecard.

7. Handle translation errors from Google and resubmit the translation, rather than just ignoring it.

As for the client-embedded translator...it's actually a good thing. There is overhead with the LSL version, but it can be just as fast as long as enough translators are worn by enough people to handle the load...but yes..we can implement more features and customize the translator for what people want. The client-embedded translator will just provide basic services. It is also the only way that IMs could be translated as LSL scripts can't access that feature.