Hi Josh, Many thanks for the clarifications.
I have done some bidirectional tests to find out how Translate handles Classical Armenian - to summarise it seems it handles the words that have the same meaning but fails on words that are spelled differently (using the classical orthography) or have a different meaning, or are not used in the modern language. It also has more trouble with sentences than with isolated words. The word order also isn't always correct. Basically it doesn't translate Classical Armenian but is relatively good at translating modern Eastern Armenian - so you may want to consider referring to it as modern or Eastern Armenian rather than just Armenian. There is an online Bible in Classical Armenian with the corresponding text of the King James Version next to the Classical Armenian verses at the following URL - it might be of help - it's not huge but it's the complete Bible and contains the main corpus of the language: http://212.34.228.170/bible_27E/index.htm I also have a list of lemmas with their English translations - please let me know if it would be of use and if so I can e-mail/upload them. Best regards e On Dec 22, 8:25 pm, Josh (Google Employee) wrote: > Hi e, > > Unfortunately it's usually impossible to tell how quickly we'll be > able to improve quality on any given language. So rest assured > quality will keep improving, but I can't guarantee how quickly, or > when we'll moveArmenianout of ALPHA. > > As for helping, generally the best thing you can do is to keep > correcting translations using the alternative translation feature we > recently launched (click on the translated phrases and select better > phrase when necessary), or if you know of huge amounts ofArmenian > data on the web, please point those out to us. Or if you somehow have > access to millions of already translatedArmeniandocuments, those can > help our system as well. > > Cheers, > Josh > > On Dec 21, 2:43 am, edd wrote: > > > Hi Josh > > > Many thanks for prompt response! > > > I do appreciate the difficulty in handling these aspects. You seem to > > be using ISO 639-1, which allocates code "hy" for all variants of > >Armenian. ISO 639-3 is more granular and has different codes for the > > different variants. > > > I guess in cases when the same word has the same meaning in all three > > variants there is no problem; but when a particular word has different > > meanings then that would affect the quality of the translation and I > > don't know how effective statistical methods would be to address this. > > Yesterday I read a couple of pages and while the translation gave a > > general idea of the text the details/specifics weren't particularly > > good and some words were just transliterated because I assume they > > could not have been translated. > > > I wonder if you could give us an idea of how closeArmenianis to > > moving from the "alpha" stage to the standard supported list and > > whether there is anythingArmenianspeakers could do to assist? > > > Best regards > > > e > > > On Dec 21, 12:34 am, Josh (Google Employee) wrote: > > > > Hi edd, > > > > Great question. Unfortunately the answer is likely "we don't", that > > > is, we don't handle the variants ofArmenian. > > > For many languages we support, there can be several variants, > > > unfortunately the difficulty for us is often that it is difficult to > > > tell the difference between documents written in the different > > > variants in an automatic way. This makes it difficult for us to > > > collect training data for our system that can separate out all the > > > variants. It is also likely that even if we could tell the difference > > > between the variants, that some variants we simply would not find > > > enough data to do anything effective with. > > > > The result is we likely find documents of all three variants > > > ofArmenianand we train our system on all of them – which probably > > > results in a system dominated by which ever variant has the most > > > content published on the web. > > > > Hope that helps your understanding, and my apologies that we cannot > > > more easily provide a way to more effectively handle these variants. > > > If you're interested in more information, check > > > outhttp://translate.google.com/about/, > > > which has a nice description of how our system works. > > > > Best, > > > Josh Estelle > > > Senior Software Engineer > > > Google Translate > > > > On Dec 20, 6:43 am, edd wrote: > > > > > Hello, > > > > I am delighted Google Translate has introduced support forArmenian. > > > > However considering there are three "variants" ofArmenian- Classical > > > >Armenian, EasternArmenianand WesternArmenian- how do you handle > > > > this "three in one" situation? Some words that are present in one > > > > variant are unknown or have different meaning in the other > > > > variant(s). > > > > ClassicalArmenianis the written language of the Bible, the church > > > > and the classical literature. > > > > EasternArmenianis the language of the Republic of Armenia. > > > > WesternArmenianis the language of the largeArmenianDiaspora. > > > > To complicate matters further EasternArmeniancan be written using > > > > the classical or the modern orthography. > > > > So my question is how do you handle all this in Google Translate under > > > > the single language code ("hy")? -- You received this message because you are subscribed to the Google Groups "General" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/google-translate-general?hl=en.
