I sent the following feedback to the G+ team but thought I’d share it to my fellow plussers in case it’s a topic you guys would like to discuss:
“Here is a feature request I would love to see implemented as part of providing a better Google+ experience.
I love to go check the What’s Hot / Explore section for new cool and interesting news. However it is often plagued by junk which has been made popular for one reason or another. I normally just mute those posts, though I will sometimes go as far as blocking a recurring poster who appear on the section but fails to provide engaging content.
What I would love is for Google to recognise that posts I mute are topics I am not interested in reading about. Similarly if I comment or +1 a lot on certain posters or topics there should be more appearing in that section or even on my stream.
I understand that the way Google Now works is to intelligently analyse our behaviour across all of Google’s services to provide an increasingly better service even when we haven’t requested it. Surely this intelligent mechanism could also be included within the Google+ experience as well (and while you’re at it, Google News too)
I know Google is always evolving and I always look forward to better upgrades to the service. Thanks :)”
CC: +Vic Gundotra +Bradley Horowitz +Natalie Villalobos +Google+ Help
Google+: View post on Google+
Post imported by Google+Blog. Created By Daniel Treadwell.
I was thinking about this right before I read your post +Jean-Loup Rebours-Smith and right after I muted another inspirational love quote post. I mute 95% of these posts that show up in my stream. Annoying that g+ doesn't recognize and take the hint. Thanks for posting this!
That sounds like a good idea. I don't use "what's hot" &c. and so it would not affect me, but if I did I would worry that posts I muted simply because there were too many replies (e.g. some Android posts, and possibly this one as well) might be falsely counted as topics I found boring.
How about Google+ filtering based on who +1'd a post, e.g. not show us posts +1'd by drooling glurge monkeys or happy clappies, and not show THEM posts +1'd by growling cynics and howling atheists.
+Martin Page What you're suggesting seems more in line with what Facebook does, ie, some arbitrary judgement of who is what based on who they interact with. I'd rather be presented content based on my habits than based on others' habits.
Yeah. But if your habits are to mute posts liked by drooling glurge merchants, then…
Hmmmm… you said: "What I would love is for Google to recognise that posts I mute are topics I am not interested in reading about." What I'm wondering is how Google would know what the topic of a post was, when they don't even seem to have the technology to translate from Spanish to English without consistently confusing subjects with objects and dropping half the negatives. If they continue to insist on doing "translation" (as they call it) without the slightest element of comprehension (they don't even bother with syntax let alone semantics) then I doubt they'll be able to know what a post you muted is about.
But they could track whose recommendations you mute.
Andrew, the translation engine is one thing, identifying patterns in posts is another, the latter would be more likely based on keywords or hashtags (when used), even identifying certain types of pictures (I systematically mute all pictures of cats for example, and I'd love for Google to do it for me at some point). Translation semantics would use a completely different algorithm, and frankly it would be handled by a completely different team / department.
I share your frustration with translation, the straight English to French translation is quite frankly appalling but that's generally due to the inherent prerequisite for context that many languages use as a base for comprehension. Translate as it is now is most useful as a straightforward phrase book but it will evolve.
Fair enough, there might be other techniques for determining the topic of posts than understanding the content, although I'd expect a lot of false positives. As for translation, you're too generous. Lack of context would be forgivable. Lack of basic understanding of grammar is not. It infuriates me that Google, Bing and other "corpus-based" engines get away with a quality of "translation" that can most charitably be described as "irresponsible", just because the vast majority of users aren't in a position to know a good translation from a bad one, and fall for the fallacy that if it makes sense it must be accurate. But anyway… that's not what you were posting about, so apologies for hijacking the thread 😉
Yes false positives would be pretty much unavoidable, which is why it should be an opt-in system (disabled by default) so you know what you're getting into by enabling it.
I would be interested in knowing more about how the translation algorithm works in detail. I know that they do use user-input for specific translations which would otherwise translate incoherently as a straight-from-dictionary phrase.
I am generous and give them the benefit of the doubt because I have studied translation and traductology at university (both French – English and Finnish / Swedish – English) and from what I recall, traductology as a science is very new and rather complex. I can't comment much about their translation to Spanish as my Spanish is still pretty rough but it's possible it has something to do with a lack of diversification. They only have "Spanish" in their list of choices, and I would assume this includes all Spanish-based south American languages. I can't imagine they don't have native Spanish / Mexican speakers amongst their staff (whether any of them works for translate is another matter), but if Spanish and Mexican is similar to French / Quebequois it's not surprising to find discrepancies (I spent far too long debating with Canadians in the past about how certain English terms and phrases should be translated in French). Again it might not be the right excuse, I'm just speculating. As it happens I ran a couple of paragraphs from a Japanese website earlier through Translate and both the English and French version were quite understandable, if not perfect.
Well, I'm not talking about dialectal subtleties here. I'm talking about absolutely fundamental, simple stuff like Google's mistranslation of a completely natural statement like "Pero no me conoce el vecino." as "But I know the neighbour" when it should be "But the neighbour doesn't know me". Took me almost no time at all to come up with such a shockingly bad example.
Also, when you say you that you "ran a couple of paragraphs from a Japanese website earlier through Translate and both the English and French version were quite understandable" what does that mean? That the result is understandable does not imply that it is an accurate translation of the original.
Funny, I'd translate it as "but I don't know the neighbour", but you probably know better 😉
Well, you'd be wrong. Of course I know better!
And what about the loss of the negative?
Regarding your second point, it very much depends on how much information you need from a text written in a foreign language. In my Japanese example it gave me information I couldn't work out just by looking at the original Japanese but the grammatical inaccuracies didn't bother me so much.
And since you seem to care so much about the Google translation engine, have you seen their "about" page? http://translate.google.com/about/intl/en_ALL/
Would it "bother" you to find out that a translation you'd trusted just because the grammar wasn't too dodgy turned out to say the exact opposite of the original?
But the problem isn't with the translation, it's with the amount of trust you put in what is ultimately an automated translation mechanism, which by definition is not going to be perfect. When you approach it knowing that it isn't going to be perfect, then it's easier to dismiss grammatical and contextual errors if the text it spurts out gives out information in an understandable manner.
Personally I'm more amused by their mistakes (which sometimes amount to detecting a completely wrong language from a website) than annoyed by it. Sure it can be annoying if you rely on such services but I only use it sporadically. Not only that, they give you the tools to help improve their service if you choose to do so.
All true, but the sophistication with which you or I personally approach (or avoid) machine translation isn't really the issue here. My point is that Google, Microsoft et al. are getting away with a shockingly, and potentially dangerously poor service because most users are not that knowledgeable and hence not in a position to evaluate it. And translators—that's actual trained people trying to make a living doing essential work—have their skills absurdly devalued as a consequence.
Actually, I should have said "all true apart from your initial statement". The problem is with the translation. I could offer an online translation service where every result was simply a random sentence created from a lexicon and a generative grammar of English, and it would fool some of the people some of the time. Would you blame them for trusting it?
This is the Internet Andrew, there are, shall we say, "less than knowledgeable" people everywhere. Google Translate is a free service, and well much as I hate to say it, you get what you pay for. If people are dumb enough to rely on free automated services, it's their loss if the results they return is ultimately wrong. If people care for full accuracy to have their documents translated, they can seek the skills of professional translators. If they don't care and publish mis-translated material which as you emphasise could prove harmful, the responsibility lies not with Google but with the irresponsible publisher. Google isn't "getting away" with a free service, they have the following disclaimer on their "about" page I linked to earlier:
"Since the translations are generated by machines, not all translation will be perfect. The more human-translated documents that Google Translate can analyse in a specific language, the better the translation quality will be. This is why translation accuracy will sometimes vary across languages."
And really, when you think about it, everything in this world is potentially dangerous.
Weeeell…. I'm not that impressed by the fact that Google has a disclaimer on its translation service. They'd be insane not to. But how many people do you think it actually puts off, in the unlikely event that they ever read it? I'd put good, solid money on a bet that the majority of people would see that and think, "OK, so occasionally the translations I get might sound awkward, a bit like something I'd get from a competent speaker who didn't have a perfect command of both languages, but maybe not quite at the level of a professional translator". They'd assume that the service was reliable (it isn't) and offered in good faith (my arse), and would almost certainly not expect something that turns "But the neighbour doesn't know me" into "But I know the neighbour". To describe that level of accuracy as "not perfect" is ludicrously misleading (and let me stress that it took me less than a minute to come up with that example).
Also, it's missing the point to characterize this as being about "dumb" people using Google to "have their documents translated". Consider a much simpler case: person A has an online friend B. B posts something on a social network in a language that A doesn't know. A runs the post through Bing or Google translate, and thinks "My God, B is a complete tosser for saying that". In fact, B said the opposite of what A thinks they said, because the translation was wrong. How much cumulative damage do you imagine is being done all the time by "little" things like that? And whose fault is it? Imagine that person A is an intelligent, sensitive person who just happens never to have learnt any foreign languages, nor to have much of a clue about how hard translation is. For example, my mother. Would you call her "dumb" for believing the online translation to be correct?
Let me make just one more point which might not be obvious: I would much rather use (and know that other people were using) an automatic translator in which the grammar of the translation itself was a bit unnatural and stilted, but which was based on syntactic and semantic analysis of the source text, and which would consequently identify and preserve subjects and objects and be very unlikely to drop important things like a negative. Such translators have been around for a long time, and reached a respectable level of competency. Corpus based translators like Google Translate and Bing do none of that. Their results can sound much slicker because they're constructed by pasting together whole phrases written by real people. And so people fall for them. The problem is that, in the absence of any true syntactic analysis, let alone the semantic kind, the pasting together is always going to be a bit like that parlour game where each player draws part of an animal then folds over the paper leaving just a bit showing for the next player to join up with. The scope for things going wrong at the joins is huge. Such "translations" are also completely biased by the grammar of either the source or target language and cannot take account of both. For example, the reason that Google Translate mangles a Spanish sentence like "me conoce el vecino" into "I know the neighbour" is that "me knows the neighbour" isn't possible with English word order. So the system has to chop the original into smaller fragments, and (I'm speculating) ends up translating "me" as "I" because somewhere else it found a translation of "me voy" as "I'm going". The fact that "me" cannot possibly be a grammatical subject in Spanish never gets a look in.
So why don't Google and Microsoft use grammatical translation rather than corpus based translation? Simple: on the surface, corpus-based translation produces results that are easier to understand, and are therefore more attractive to the user. That those results are not actually an accurate translation of the original is, fortunately for Google and Microsoft, but rather unfortunately for the users, completely hidden from view.
I would never dare place such a judgement on your mother Andrew. Having said that, if my parents were running my posts in English through a translation service and build up a false assumption of what I'm saying because they did not ask me to clarify in a language we both understand it may be a different outcome. I am certainly not responsible for anyone's misunderstandings.
And, well you do make a lot of negative assumptions. I'm a bit at a loss about why you feel so strongly about this. To take your social network example, given the choice between having a translation service and not having one, given that the former choice saves you having to write entries in multiple languages, what would you choose? I'm not judging your choices in the matter (please don't make that assumption) but for my part, if I have to say something of importance to people in both languages, I will go through the trouble of writing in the two languages, as I'm sure you would.
I've seen my share of people jumping to conclusions far FAR too quickly because they misunderstood something and didn't bother seeking clarification (usually in the same language). If someone who follows me but who doesn't speak English and judges me harshly due to their relying on a third party to translate things for them, I'm not going to feel responsible for their decision to place a certain judgement.
You have every right to be cynical. When we come to rely on certain services and they fail us repeatedly (I could give you a rant bigger than yours about Facebook, believe me), we lose trust in its efficacy and usefulness. In the end it's all down to what our relationship with the service is, not the service itself.
On your second point, yes you may be right about their reasons being producing results which are easier to understand but my guess would be in terms of computational value, adding complex (I'm thinking French here ok ;)) grammatical and semantic rules to a system which focuses on being as instant as possible would perhaps be counter-productive.
Aka, that's a lot of if statements to put in there 🙂
… not to mention the exceptions
Apologies for the hiatus, which was due to an overnight change of altitude. But not attitude, alas 😉 Just a couple of brief answers anyway. First, I'd rather that people had no translation service than a completely broken one. And second, instant gratification is overrated: the argument that poor quality results are justified because they don't take long to compute doesn't convince me at all.
Finally… is this really Jean-Loup Rebours-Smith saying "it's dumb to trust free stuff" or have you been replaced by an impostor? 😉
(To clarify: quickly-computed but rough, imprecise, low-quality output that you can see is of low quality is entirely defensible. The problem, as I hope I've made clear by now, is when you have no way of knowing that the quality is low).
No I'm not saying it's "dumb" to trust free stuff, I'm saying it's illogical to fully rely on a machine translation software for your translation needs. It has nothing to do with free vs licenced.
And, even if it's done by a professional translator, if you have no knowledge of the source language, how would you know whether the quality of the translation is high or low? With a flesh and blood translator you trust their expertise and experience to deliver but humans aren't infallible. Where do you draw the line of trust? It's all subjective in the end
(sorry for the incoherent thoughts, I realise there are inconsitencies in my arguments but I'm home alone with Emily and she's demanding my time 🙂
With a flesh and blood translator I'd ask to see their qualifications.