3 ms·
> Google goes via English. No, it doesn't. It goes via an intermediate representation which is common for all the languages they translate. The meaning of a ph
by visarga 9y ago
> Google goes via English.
No, it doesn't. It goes via an intermediate representation which is common for all the languages they translate. The meaning of a phrase is represented as a "meaning vector", or more recently, a matrix. The problem is that it is hard to translate rare words, because when the neural net decodes the next word, it needs to consider the whole dictionary at once (the problem of too large softmax function slowing down the process). So they need to focus on a subset of words. There are various solutions, such as hierarchical softmax, but they are still difficult to train. Google translate uses a ton of compute to create a model as is. They probably can't offer everyone a version that is 2% better but uses 100x more resources.
- pawelk 9y agoHow do you get from Polish "hebel" to German "Flugzeug", if not by dropping the "hand" part of "hand plane"? Especially when the Polish is a loanword from German Hobel...
- visarga 9y agoWe've become too entitled and can't appreciate the huge amount of work that went into automatic translation any more. If GT doesn't cut it, we can always take a ... you know ... dictionary and do it by hand. Google search isn't perfect either, Wikipedia has errors, Twitter has bots, why are we bothering to use them? Because they are better than not having them.
- pawelk 9y agoOr maybe one could reasonably expect that Google, with all that money and engineering power, would implement something like a dictionary lookup to short-circuit single word translations? Edit: I did not want to complain about the translation quality nor show my sense of entitlement. You have authoritatively stated and used a some jargon to tell us it doesn't use English when translating between languages, so I simply wanted to understand more about this process because, on the surface, it looks like English was involved in my specific example.
- dewey 9y agoDon't you think that if it would be that easy they would've already done it?
- pawelk 9y agoI don't say it's easy, at least from organizational/product-development perspective. But that's basically how every translation website before GT worked: type a word in language A, select language B, get the result(s) from a dictionary, done. In the nineties I had this pocket device: https://i.imgur.com/Yp9yTRL.png https://i.imgur.com/Yp9yTRL.png which would look up and translate words in 8 languages. Today I have a smartphone which can connect to a datacenter with enormous computing power, but a dictionary lookup is a hard problem?
- umanwizard 9y agoThere are plenty of dictionary lookup apps, if that’s what you want. It’s just not what GT is.
- pawelk 9y agoAs a developer I kind of get it: "we have all this machinery so we won't compromise to use the simple route even if it makes sense short-term; let's improve our product instead". But as a user I'd expect Google to be at least as good as dictionary lookup from two decades ago, and do PL -> RU or PL -> DE instead of PL -> EN -> RU, PL -> EN -> DE when I type in a single word.
- umanwizard 9y agoGoogle Translate simply isn’t meant to be used for dictionary lookup, though. If you use it for that then you’re using the wrong tool. This is like complaining that Word isn’t as good as vi, which is decades old, for text editing. Sure they could bolt a text editor onto Word if they wanted to, but why bother? That’s not what people are looking for when they use the product. And if that’s what you want, there are many other choices that are much better.
- pawelk 9y agoPolish PM would probably disagree it's better to have imperfect auto-translation than none at all ;) https://www.bloomberg.com/news/articles/2018-02-02/translation-mix-up-dents-polish-efforts-to-defend-holocaust-bill https://www.bloomberg.com/news/articles/2018-02-02/translati...
- intopieces 9y agoDissatisfaction with the status quo is precisely what keeps automatic translation companies in business, and indeed drives all innovation. The author explicitly gives credit to GT for being good enough for lots of people most of the time, but that does not excuse its fundamental failures. GT Engineers don't want a pat on the back. They want to do better.
- majormajor 9y agoLots of discussion about compute-intensive alternative solutions there, but what about a human layer where something triggers a "wow, this word is rare, let's call up Prof Whoever and ask them for a suggestion" notification? There's a lot of words in the world, sure, but you only need to solve each one so many times. Dictionaries seemed to manage this trick even before computers! Similarly, I would pay a lot of money for a Spotify or Netflix alternative where a knowledgeable human looked at my activity every month or so and created a hand-tailored recommendation list, since those products seem pretty bad at nuance in ways that even just simple "do you like this for reason A or reason B" questions would answer. I know how the sausage is made at some non-Google places, and there's a common institutional blind-spot where the algorithms people don't want any human-driven chocolate getting in their peanut butter, and lazy product people are perfectly content to let any junk the algorithm throws up be the math guys' fault. Sadly, the world is full of edge cases, and sometimes you need one-off rules. (There's a parallel thread going below where people are aggressively defending the fact that the translate product as implemented is bad at certain use cases. Baffling. At the end of the day, are you happier telling users they're doing it wrong than given them what they want?)
- YeGoblynQueenne 9y agoYes, I know about GT's "interlingua". I've chosen my "swallow" example specifically to show that it's basically just a model of English. Which makes a lot of sense when you realise that all the languages in the language pairs that GT's interlingua covers will have many, many more texts going to and from English, than to and from any other language. Attempting to derive a many-to-many language model from an English-dominated corpus is very likely to lead to an intermediary model that is basically English-to-many. In fact, if you read the relevant paper [1] the dominance of English in tables etc jumps out at you. And if you look at how Google figured out its Neural Machine Translation system has created an interlingua, that is ... visualisation. So it's not like they have some kind of principled way to demonstrate that this interlingua they advertise actually exists. They just look at some pretty images and interpret them to say that it does. Even this visual "evidence" as they call it, is from a many-to-many language model trained on Japanese, Korean and -you guessed it- English. __________________ [1] https://arxiv.org/pdf/1611.04558v1.pdf https://arxiv.org/pdf/1611.04558v1.pdf