When I was at school (in the 90s) we were still getting taught French with textbooks that had such key phrases as "there's too much blue in my laundry" or "you don't put enough starch in my collars". I suspect these textbooks had been written by Enid Blyton.
Loose tongues and oily seamen: Lost in machine translation yet again
My uncle has an airship. Correct me if I'm wrong but this language course seems a little old-fashioned. Mother orders Cook to buy more servants. I should have guessed something was up at the library desk when they mumbled something about having to "fetch the discs". While I was still racking my brains as to where my last …
COMMENTS
-
-
Friday 12th July 2019 09:54 GMT I ain't Spartacus
We had Jean-Paul and Marie going to the le discotheque. In their best polo neck jumpers - though at least they weren't smoking Gauloises. I mean, who goes to the disco with their sister anyway? Leaving aside that I bet it's been le nightclub for quite a long time.
At least I never came across the phrase, "pass me my chest wig and gold medallion..."
-
Friday 12th July 2019 10:16 GMT Frederic Bloggs
When I was at school (in the 60s) my mother regularly used to add a small fabric covered blue coloured "pill" (about the diameter of a two pence piece and 1cm thick) which was partially dissolved in the wash to make whites look whiter. This was superceded by the likes of Daz (and other brands) containing "blue whiteners" at about the same time.
Mind you she was never one for using starch.
-
Friday 12th July 2019 17:18 GMT Antron Argaiv
We had Monsieur et Madame Thibaut et ses enfants Paul et Catherine. Monsieur Thibaut est ingenieur.
More importantly, they drove a voiture, a word I never heard when I visited Paris. So, 10 years out of date, perhaps. I don't recall a TV in their house.
Filmstrips and an audio tape. Ecoutez, et repetez!
Having spent 3 years in French-peaking Brussels as a tike probably helped a lot. I took French classes from 6th grade through my first year at university. I can still read it and understand it 50 years later.
-
-
-
-
-
-
-
Friday 12th July 2019 19:41 GMT Rich 11
Re: bread is bad for ducks.
I'm gratified to see that a number of people share my concern for the well-being of Anatidae, existent or non-existent.
Next week: Would you allow Mustelidae to share your child's bedroom?
-
-
-
-
This post has been deleted by its author
-
-
-
-
-
-
Saturday 13th July 2019 17:33 GMT Anonymous Coward
Is dat neet "in ut plat"?
Het hungt er auch nog vanaaf aan welke kajnt van de grens se bès.
(sorry guys, we've wandered into a language exchange that somehow strayed into the only European language (yes, that's official) where tonal elements can change the meaning of some words. Just pitch accents indicating a variation, not the complete word changes in languages like Chinese, but it's unique in Europe).
Regular programming and sarcasm will now resume.
-
-
-
-
-
-
Friday 12th July 2019 12:33 GMT Alien8n
Re: Ah, French, that's easy
My latest music festival jaunt included an interview in Mongolian.
My colleague however managed to confuse Korea with Nepal, but this didn't matter as the Nepalese singer of the band I was given to interview actually spoke better English than my colleague.
He then had the pleasure of mangling Japanese.
The New Zealand band that sings in Maori were fun though.
-
-
-
-
Saturday 13th July 2019 17:43 GMT Anonymous Coward
Re: Ah, French, that's easy
I have a friend who is half Swiss, half French. The Swiss part shows by her being very detailed and reliable, the French part by insisting that Parisien French is the only proper French, and it has to be spoken a certain way (she was brought up there).
That said, her English is very, very good, no need for me to struggle with my French.
-
-
-
-
-
Saturday 13th July 2019 17:37 GMT Anonymous Coward
Re: Ah, French, that's easy
Now Swahili and Mandarin pose a bit of a challenge...
In a job I am not allowed to name I made that joke once when we had some garbled bit of text: "as far as I can tell it's in Swahili", only to be corrected by the guy who ran the unit as he turned out to be the only person I have ever come across that actually speaks it. So, garbage it was :)
-
-
Friday 12th July 2019 11:08 GMT Anonymous Coward
Simultaneous translators have problems with German where speakers the verb at the end often and without consideration for the translator put.
But my favourite simultaneous translation story relates to a conference at which a Bulgarian speaker got up and launched into a diatribe full of Marxist-Leninist jargon. Other translators struggled through, but the French translator simply repeated over and over again "French is a precise language and has no way of conveying what the speaker is saying."
I have heard it suggested (by a French engineer, of course) that if a business expression cannot be converted into immediately clear French, then it is not worth the bother.
-
Saturday 13th July 2019 17:55 GMT Anonymous Coward
Try the German term "herkömmlich". What makes it hard is not the factual meaning which is "usual" as in "the usual products", but conveying the associated negative connotation, so it's more "the usual products which are nowhere near as good, reliable (etc)".
It's a spectacular word for marketing because it's a one word equivalent to a paragraph worth of loathing :).
-
Friday 12th July 2019 09:19 GMT macjules
Cryto Lotto - Watta Lotta Crappa
I feel it is only right and proper that I set up a comparable service for Lotto
suckerspotential billionaires to be able to redeem their inherited goods and gold. I shall therefore create an online portal called "We Buy EVERYTHING of yours" such as your car, your children's university trust fund, that little nest egg that you thought you could hide. In return for our carefully audited appraisal (we have engaged a leading auditor in London) we will give you 1 bitcoin for everything you own, which you can then spend on the Crypto Lotto, since you are bound to win anyway.Remember the value of bitcoin can go up as well as plummet to new unforeseen depths.
-
-
Sunday 14th July 2019 02:48 GMT stuartnz
Re: Deepl >> Google Translate
Thanks for the tip! I've never had much of an issue with Google Translate (which I have long called Goofle) handling European languages, but its Hindi was risible, and its Korean was basically random strings that made graunching marmosets seem like Shakespeare's finest work. Both are better now (setting the bar low, ofc), but if anyone is in need of a passable machine translator from Korean into Englidh (can't vouch for it the other way) https://papago.naver.com/#/ko/en is better than Goofle by an order of magnitude. Especially with cultural references and some of the impenetrable portmanteau coinages that colloquial Korean is obsessed with
-
-
-
Friday 12th July 2019 09:50 GMT Stumpy
Re: Prior art
Dutch: Mijn hovercraft zit vol palingen
Basque: Nire aerolabaingailua aingirez beteta dago
Czech: Moje vznášedlo je plné úhořů
French: Mon aéroglisseur est plein d'anguilles
German: Mein Hovercraft ist voller Aale
And, of course, in the original Hungarian: A légpárnás hajóm tele van angolnákkal
-
Friday 12th July 2019 15:20 GMT lee harvey osmond
Re: Prior art
Yep. Demonstrated that to our French visitors during the twinning society visit last month.
I was sure there’d be at least one devout Orthodox Pythonist among them. It seems there wasn’t.
And I listened to a lengthy complaint in French from one of our guests, who had an iPad with the same app I was using, except it appeared she hadn’t followed the recommendation to download the required language packs, and her iPad was a WiFi-only model.
En principe, oui, le singe est dans l’arbre
-
-
Friday 12th July 2019 09:32 GMT Kubla Cant
Mon oncle a un dirigeable
Yes, yes. But has your postillion been struck by lightning?
I was delighted to discover that this isn't entirely apocryphal. According to a reliable source, John Murray's Handbook of Travel-Talk (1847) includes this calamitous scenario:
Oh, dear! The postilion has been thrown (off) down.
Is he hurt? Run for assistance to the next cottage.
Ask for a surgeon.
I am afraid he has broken his leg — his arm.
He has been bruised on his head.
He must be carried home gently.
It rains in torrents.
It lightens—it thunders.
-
Friday 12th July 2019 09:48 GMT Potemkine!
Brian is listening to music on Radio Blackpool
At least that's what we were told when we were beginning to learn English.. We also learnt that Brian was often in the kitchen, when Jenny was in the bathroom. Go figure....
Google Translate is not that bad. Every day we have many lorries coming from all over Europe to pick or deliver merchandise. Lorry drivers come often from eastern Europe and don't speak a word neither in French nor in English. So we equipped our logistics team with an android tablet with Google Translate on it. After several months, experience shows that it quite efficient to make communication possible between our team and the drivers, and the logistics team is delighted.
-
Friday 12th July 2019 11:18 GMT Anonymous Coward
Re: Brian is listening to music on Radio Blackpool
Google translate works absolutely fine with most languages so long as you start off by expressing yourself clearly and without jargon. The problem is that many people are quite unaware when they are using jargon, and are surprised by the result.
I remember reading that The Sun employs English graduates as sub-editors because they are able to express ideas at the 12-year-old reading comprehension level for which it aims. You have to know a lot to write simply.
-
Friday 12th July 2019 18:59 GMT Ken Hagan
Re: Brian is listening to music on Radio Blackpool
"I remember reading that The Sun employs English graduates as sub-editors because they are able to express ideas at the 12-year-old reading comprehension level for which it aims. You have to know a lot to write simply."
I have heard something very similar from a relative who had been employed in that (or very similar) capacity by The Sun.
The great pity is that we were all 12 once and somehow most of us appear to have lost this skill.
-
Friday 12th July 2019 22:00 GMT stiine
Re: Brian is listening to music on Radio Blackpool
I misunderstood the parent post. I made the assumption that The English Graduates were only capable of reading and writing on a 12th grade level.
Personally, i was reading on the 12th grade level when I was in 5th grade. Most likely due to the fact that I read a 1960's Colllier's Encyclopedia from Aardvark to Zwingli (minus 2 volumes) for fun during summer breaks between 3rd and 5th grade.
-
-
-
Friday 12th July 2019 12:36 GMT jrd
Re: Brian is listening to music on Radio Blackpool
I suspect that if you know you are going to be speaking through a translation program, it is relatively easy to use simple sentences and avoid words and sentence structures which are likely to be troublesome. Which is pretty much what you'd traditionally do when speaking with someone not fluent in the language.
-
Friday 12th July 2019 18:25 GMT FrogsAndChips
Re: Brian is listening to music on Radio Blackpool
Never learnt English with Brian but it seems every other Frenchman did. Our main character was a young bloke living in Wimbledon but who didn't like tennis. Can't recall his name, but 30+ years later I can still remember his phone number. "Hello, 6234789?"
-
-
-
-
Friday 12th July 2019 11:05 GMT Mage
Re: Try Irish names
Even if you hear them spoken, even in Ireland you ask how it is spelled, because there isn't one spelling for many "traditional names". But forget spelling properly for the Irish Health service which obviously uses 1970s software, can't do á é í ó ú Á É Í Ó Ú. Many names spelled without the accent (fada) are completely different words with different pronunciation.
Google Translate is pretty garbage for Irish or to send to a native speaker of a language you don't know. It's not too bad for your own use. It's dumb pattern matching from OCR text of books and documents. Maybe when Google illegally scans the Irish Harry Potter Translation and compares it to English it might improve. Mostly for Irish they have EU texts. That's what they started with.
-
-
Friday 12th July 2019 13:32 GMT Loyal Commenter
Re: Try Irish names
If you think about it, it is mildly interesting that the translations of Asterix from French into English involved a lot more creativity than a straight word-for-word transliteration. I can't remember where, but I read an article somewhere where the woman who did the translations was interviewed, and it involved coming up with new names which were puns that worked for several of the characters, especially where the word in French is substantially different from the English. The fact that the transaltion works at all in many places is down to the fact that a lot of words in English or French have been borrowed from the other. I wonder if a meaningful transaltion could be made at all in irish Gaelic.
-
-
Sunday 14th July 2019 20:56 GMT Dave559
ASCII
Re: á é í ó ú Á É Í Ó Ú, etc
It has always struck me as a particular irony that the country which probably has the greatest mix of original nationalities and names in the world was the one that gave us the very restricted character set of ASCII, so that many of its own citizens wouldn't even be able to type their own names properly. Although many of them had already had their names clobbered by the immigration officials at Ellis Island first, sadly.
(And, yes, I realise that ASCII was invented back in the days when they couldn't even spare 4 digits for the year, a problem which also took quite some time to be resolved.)
-
Sunday 14th July 2019 22:21 GMT jake
Re: ASCII
Well, to be fair, ASCII (more properly US-ASCII) came about from the widespread use of the English Alphabet, which has it's modern roots in the work of Byrhtferð back in the year 1010ish. Back then, the only European style writing in North America would have been Runes. Blaming "the country which probably has the greatest mix of original nationalities and names in the world" on the accident of history which lead to modern English orthography seems a trifle puerile.
-
Monday 15th July 2019 09:40 GMT Anonymous Coward
Re: ASCII
the country which probably has the greatest mix of original nationalities and names in the world was the one that gave us the very restricted character set of ASCII
Isn't that sort of inevitable? When you have a wide range of data that you need to fit into some common, broadly-usable, form, the general approach is to drop the outliers and focus on the lowest-common denominator. It's not ideal for anyone, and loses the subtleties that can really make a difference, but works after a fashion for them all.
Think of it as the EU approach to unity... OK, OK, I'm leaving...
-
-
-
-
-
-
-
-
-
-
Saturday 13th July 2019 13:36 GMT MonkeyCee
Re: Mijn grootvader is een schaap
Based on the attitude of my male Dutch classmates, some of them genuinely believe that a) English has a correct form and b) that is what is taught in Dutch schools.
Explaining why using Dutch/German sentence structure, then re-arranging it with a dozen commas per sentence doesn't look right, and having the blokes correct me was pretty funny. Had the lecturer come over, listen for a moment, and say "Would you trust an Englishman to tell you how to write Dutch?".
I've also joked that I speak about five versions of English, British home countries, upper class USA, working class USA, Indian trade English and antipodean potty mouth.
Groot vader? Weird. Hardly ever heard vadar (my brat calls me papa or dad, depending on language) maybe in more formal situations. Certainly never heard groot vader, only opa :)
-
-
-
-
-
-
-
Friday 12th July 2019 10:12 GMT I ain't Spartacus
At least vinyl has made a comeback
What with records being fashionable again, at least there's a chance you can play your language course. I was going though my old box of computer crap the other day, Mum needed a spare power cord for a laptop. Turns out I've only got the old style kettle plug type, and not the newer 3 circles ones - and I realised how old my collection of "useful" stuff is. Too many parallel printer cables, and my copy of Elite on 3.5" floppy for DOS 5.0.
I also found my french language course from when I moved to Belgium. It's on cassette tape. With what shall I play them? More stuff for the bin, so at least the spare room is getting cleared.
Oh and Monsieur Dabbs - thanks for the term "wanker cash". When these people tell me that crypto-currency is the future, when fiat currencies collapse - I can simply reply, "your wanker cash is rather more likely to collapse first."
-
Friday 12th July 2019 11:34 GMT Anonymous Coward
Re: At least vinyl has made a comeback
"my copy of Elite on 3.5" floppy for DOS 5.0."
Odd, I thought my copy of DOS 6 came on 5.25" floppy. I brought some 5.25" floppies into work yesterday for engineer show+tell; at least one of my co-workers had never seen one. A whopping 360 kb each!
"It's on cassette tape. With what shall I play them?"
I have a nice dual cassette deck, where each side can play both sides of the tape. (Think party - pop in 2 tapes, it plays front and back of each tape in sequence.) I hooked it up to my computer, and spent the past couple of weeks transferring tapes, mostly audiobooks. I'd better hurry - I think the belts in the deck are breaking. One side of the deck will no longer play either direction, and the other side only plays side B, not A...
-
-
Friday 12th July 2019 12:59 GMT I ain't Spartacus
Re: At least vinyl has made a comeback
Oh, I remember him now. Wanker Cash was great on radio!
'Ee sang so many great songs:
I Walked the Shoreline
Tram of Love
Rock Island Line
And of course - The Man in Blackpool
My coat? The one with the little stick of Blackpool rock in the pocket please luv.
-
-
Friday 12th July 2019 13:11 GMT Anonymous Coward
Re: At least vinyl has made a comeback
It's still very easy to get cassette players - I bought a language lab one in full working order for £25 - and the wonderful analog output means you can transfer the tapes to computer if you wish. Then sell the recorder again on eBay.
When I digitised all my slides, I was able to sell the scanner for nearly what I paid for it, saving me hundreds (if not thousands, they were 645 and 6x6) of pounds that a bureau would have charged.
-
-
Friday 12th July 2019 10:54 GMT Mage
are better understood these days
No, they are not. There is just a bigger pattern matching database. No understanding from context. Using all those EU documents and different language editions of books.
Don't look at the man behind the curtain, possibly illegally scanning texts from libraries.
-
-
Sunday 14th July 2019 19:08 GMT jake
Re: are better understood these days
"It is perfectly legal to scan library texts
::snippage has occured::
What is not allowed is to reprint and sell copyrighted material."
Please note that this depends on jurisdiction. The gory details of copyright law vary widely from country to country. Caveat auditor.
-
-
-
Friday 12th July 2019 11:06 GMT Anonymous Coward
You underestimate the rate of progress,
I think.
As a translator, I've been watching google progress on and off over the last, what, 10 - 15 years. And, while I remember correctly, we were all having a jolly good time sharing "those" memorable google translations in those days. Truly hilarious, that google translate, I tell you. Then, the novelty disappeared I suppose, and we have got back to work and worrying about dwindling translation rates. Every now and then google translate would come up again with something extraordinary, worth sharing and wasting a minute or so, but that was it.
...
No, there's no mega-ueber-punchline, it's just that this year, I've put some texts, around 4 - 5 in total, in Russian, Polish and German into google translate, to share some movie reviews and lighter journalism pieces with people who only knew their native tongue, and I was [unpleasantly] surprised to find those translations came out either near-perfect or very good (with very few glaring, easily fixable issues). And today I've put a German bank account application through google translate too (because I'm a lazy bastard), sentence by sentence, and it is reads reasonably well in English.
On top of that, a few of my travelling friends have mentioned some translation gadgets / app combinations that let them speak and out comes the translation in the natives' tongue (we come in peace, where do you hide bitcoins / women / wine?, etc). I can't verify the quality of translation, because those apps always want internet connection and they connect to god-knows-what (plus google, of course), so I'm not having any of that. But people are happy enough.
So, looking at the developments, perhaps that combination of monkeys hitting them typewriter keys long enough to come up with Shakespeare isn't that far from the truth. Sure, you might get Shakspeer, but this is probably more than good enough for HUGE majority of "consumers".
Oh, and by the way, apparently a current trend in university translation courses is editing. Post-translation-software editing. Time to die. Or re-train as a word-processing clerk...
-
Friday 12th July 2019 11:38 GMT Anonymous Coward
Re: You underestimate the rate of progress,
I was recently given some Hungarian coffee. I only speak English. I took a photo of the package and ran it through Google Translate. I can't be completely certain, since I can't read the original Hungarian, but I'm fairly confident the text does NOT say that the coffee is "pork-free" and has a "fashion disaster", which is part of what Google produced.
-
Saturday 13th July 2019 13:45 GMT MonkeyCee
Re: You underestimate the rate of progress,
"those translations came out either near-perfect or very good (with very few glaring, easily fixable issues)."
For whatever reason google translate gets negatives (or double negatives) wrong in Dutch. So you get things like "off peak tickets may not be used after 9am or at all on weekends and public holidays".
Usually gets it perfect for stuff from the tax department. I assume it gets lots of practise on those :)
-
-
Friday 12th July 2019 11:45 GMT Paul Cooper
Lost in translation
I am married to a chinese lady, and often have to resort to Google translate when participating in WhatsApp conversation with her relatives. The results vary between being a useful guide to what was actually said and something from Monty Python. Google Translate doesn't cope well with Chinese! Fortunately my wife can intervene if I get the wrong end of whatever is being said, and my relatives tolerate my lack of understanding of their language.
I also revise English written by Chinese speaking academics.
Both are very interesting as an illustration of one of the major problems with translation, and that is that the underlying assumptions of a language can be wildly different. My academic clients make regular mistakes that arise from the totally different logical basis of chinese languages (yes, there are several and they aren't mutually comprehensible). For example, Chinese languages don't have time binding, so most of my clients apply the tenses of English without clear understanding of what's going on, with results that vary from strange to incomprehensible! And they mix colloquial English - even slang usages - with excessively formal constructions. ENglish has a clearer distinction between written and spoken language than Chinese does. And do it goes on.
Machine translation is never going to be really effective until it can take into account the different cultural bases of language, and the very different ways language can be put together. Indo-European languages are unlike Chinese languages, and no doubt the same goes for other language groups that I am not familiar with.
-
Saturday 13th July 2019 15:23 GMT Barry Rueger
Re: Lost in translation
We took an introductory Mandarin course this year, and I honestly loved the underpinnings of the language, especially the lack of "irregular" verbs. Even the tones were less challenging than I expected.
Sadly the actual content of the course was entirely useless for tourist travel. The sentence that still sticks with me is:
我姐姐会喝杯柠檬茶
Since I lack an older sister, and don't drink lemon tea....
(Huawei phone translation, no guarantee it's accurate.)
-
-
Friday 12th July 2019 12:19 GMT Chairman of the Bored
The human Rosetta Stone
Sometimes you don't need a gadget, you need a guy.
Like it or not, in the US most dirty dangerous construction work is done by people of questionable immigration status.
I had a new roof put on, and the workmen accidentally dropped about a ton of crap onto a brand new HVAC system. Instant destruction. Suddenly, none of the men could speak English. Furthermore, they pretended they didn't understand my Spanish.
Crusty old retired Navy Chief comes over, stares at me, and says, "[Chairman], dumbass, you're not communicatin''"
Me: "Oh, FFS. What do you expect me to do?"
Chief, to foreman: "If El Manajero ain't here in five minutes, I'm callin' Immigration. INS! I bet evry' swingin' dick here is back on the boat by Saturday!"
Me: "..!"
Suddenly everyone spoke fluent English. I saw El Manajero is less than five, and had the insurance firm on the phone in maybe ten.
-
Friday 12th July 2019 13:51 GMT Stork
The hardest about languages
The hardest is when you can make a meaningful translation without translation the meaning.
The Portuguese really do not like to say no. An awful lot of them speak _very_ reasonable English, but "it's difficult" or "it's complicated" is not quite the same as "no".
-
Friday 12th July 2019 14:10 GMT tony2heads
You have encapsulated my world view
'I have the opposite problem: the outside world to me is a foreign country filled with weird people saying and doing weird things that defy explanation or logical purpose"
So much of what passes for popular culture and politics are covered by that short comment.
-
-
Saturday 13th July 2019 12:49 GMT Fruit and Nutcase
Re: Banning cryptocurencies on environmental grounds?
Porn produces same amount of carbon dioxide as whole of Belgium, study finds
-
-
Friday 12th July 2019 16:06 GMT onemark03
Dear Mr Dabbs,
As a professional translator from German to English, I can warmly recommend the DeepL machine translation tool. It's not perfect and is about 95% effective but is getting better all the time. You do have to check your work, but you would do that anyway.
It also works both ways.
Yours,
onemark.
To Grumpenkraut:
The German translation of paradigm shift is "Paradigmenwechsel" (= "paradigm change").
Just saying.
-
Friday 12th July 2019 18:38 GMT Blackjack
Robots stole your piggy bank
More and more tasks humans used to do are being done by tobots.
The 2019 version of Child's Play horror movie is basically a smart home going wrong and ending with a talking doll version of Alexa killing people.
Digital malware can already kill people by screwing up with hospitals equipment. Badly coded self driving in cars has and is killing people while autopilot in planes has killed people because it lacks an easy way to disable it.
Do we have to worry about our smart homes ordering a hit on us?
Maybe that new smart seatbelt in our car might get a tad too tight?
Will our fitness tracker tattle our location to home burglars?
The future is here, and the robots are not only stealing our jobs but also killing us. Oh and stealing our data (and other things) from us too!
-
Saturday 13th July 2019 01:40 GMT Michael Maxwell
"ordering another 10,000 toilet rolls" You laugh, but in The Ship with the Flat Tire (https://www.goodreads.com/en/book/show/18001514-the-ship-with-the-flat-tire), the Supply Officer (a fresh Ensign) orders a certain quantity of TP, because his Chief thinks the unit of issue is the roll. Turns out the unit is the case--something like 100 rolls per case. When it arrives, they have trouble finding places on board to stow it. But as you might guess, it saves the ship later on.
-
-
Monday 15th July 2019 21:04 GMT I.Geller
I mean the following: there is a sentence "Alice and Greg swim with joy." If Google doesn't see each word's part of speech, then the word 'joy' can be taken (by computer) as a noun (name), resulting in erroneous patterns when parsing the sentence.
If the word "joy" is a noun, then these patterns appear:
- Alice swims
- Greg's swims
- joy swims.
If the word "joy" - an adjective, then these:
- Alice swims with joy
- Greg swims with joy.
People can immediately understand what the word "joy" is an adjective because it doesn't begin with a capital letter. But computer must be pre-programmed accordingly. And if not?
If not there are errors and, as a result, these terrible translations: this sentence can be found then a search request "Does Joy swim?" used, even if it has nothing about Joy. Words with wrong parts of speech potentially can become the source of lexical noise, erroneous patterns and wrong translations.
This post is translated from Russian, using my skills and Google translate.
-
-
Tuesday 16th July 2019 00:27 GMT I.Geller
Re: Just to further make people think ...
For example, anther sentence: “The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.” If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators.
Thus some patterns are lexical noise and should be somehow either deleted or ignored. The Microsoft team did that using a new method based on the semantic similarity between the pronoun and its antecedent candidates, which I patented as surrounding paragraphs's patterns. As the result Microsoft has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall average score on GLUE (87.6 vs. 87.1) on June 6, 2019.
As for Joy... You never know what programmers can invent drinking a beer, this is the reason quality control exists.
-
Tuesday 16th July 2019 00:27 GMT I.Geller
Re: Just to further make people think ...
Is it possible that the computer sees "j" and "J" as the same thing? No matter how unlikely but it's possible, isn't it?
There is no room in SQL database for possibilities and probabilities, only for the certainty. The AI database is no different, all lexical noise should be purged.
I've personally seen a few cases where capital letters were ignored and all words were treated as starting with a small letter, no matter what letter they really started with, because it didn't matter.
-
-
-