Show HN: Master Numbers in a Foreign Language | Hacker News Reader

Show HN: Master Numbers in a Foreign Language

foreignnumbers.com

112 pointsjasonlfunk80 comments

For anyone who hasn't studied a foreign language extensively, you might be surprised to learn that numbers are much more difficult than your intuition would suggest.

You might think, what's the big deal? You need memorize around 40 words (i.e., the words for 0-9, 10-19, 20, 30, ..., 90, 100, 1000, million, etc.), plus a few rules, and then practice until you're fast.

I want to point out some difficulties that people learning English encounter. Every foreign language will have its own anomalies.

- In English, a year like 1776 is spoken as 17-76, not as one thousand seven hundred seventy six. And if you say it the long way, a native English speaker might even get a little thrown off.

- As an exception to the above, a year like 2001, is spoken as two thousand and one, not as 20-01.

- As an exception to the exception above, a year like 1701 is spoken as 17-oh-1, and not the long way.

- Times like 3:45 are spoken in numerous ways, like 3-45, quarter to 4, 15-45 if afternoon, or possibly even 15 to 4.

- When giving an address, 386 Maple is said as 3-86 Maple, not as 3-8-6 nor the long the way.

- You need to learn fractions and ordinal numbers: half, one-third, an-eighth, etc. You eat a third of a cake on the third floor, but half a cake on the second floor; i.e., different words in the case of number 2 (half vs second).

- When speaking a series of digits, 0 can be said as oh or zero; in many contexts, oh is much more common.

- Other exceptions and special cases with dates, prices, decimal points, measurements, ...

So it's not just programmers that have a hard time with dates/times, it's the entire human race.

This is great. Since we speak the language every day (even though I'm not a native speaker), we do not see the minute details of it.

I could think of one more. 308 - Lot of people pronounce it as 3-naught-8.

When we see a score, 1-0, we say 1 - NIL.

0 is complicated.

https://en.wikipedia.org/wiki/Names_for_the_number_0_in_Engl...

Those examples also show regional variation: in most forms of U.S. English "naught" and "nil" are almost unheard of today and those examples would be "3-oh-8" and "1-zero" (or "1-nothing" or "1-oh", or maybe some other variants that would probably be marked as humorous or satirical like "zip" or "zilch").

True but when we talk about nuances in English numbers, we do have to consider the other common ways. For someone who is learning English numbers and watches, say football (soccer), they would often hear 1-nil in commentary. If they are restricting themselves to just U.S. English numbers, they would be left confused in this example.

Yep, I meant that such regional variations mean that we have even more to learn about English numbers and make them even trickier!

Check out the awesome rules for Arabic:

https://en.wikipedia.org/wiki/Arabic_grammar#Numerals

> Numbers behave in a quite complicated fashion. wāḥid- "one" and ithnān- "two" are adjectives, following the noun and agreeing with it. thalāthat- "three" through ‘asharat- "ten" require a following noun in the genitive plural, but disagree with the noun in gender, while taking the case required by the surrounding syntax. aḥada ‘asharah "eleven" through tis‘ata ‘asharah "nineteen" require a following noun in the accusative singular, agree with the noun in gender, and are invariable for case, except for ithnā ‘asharah/ithnay ‘ashara "twelve".

Note that this is just the introduction and doesn't purport to set out all of the rules... (!)

This is great feedback!

In addition to this, I think numbers really help with the soft skills of listening and producing - since the actual number of words is rather small but you can mix and match them in an unlimited number of ways. You can't just understand most of it and flub your way through the rest.

good ones! as an english speaker, some things that gave me difficulty in a second language (see if you can guess it):

- some digits are pronounced differently for counting than as part of a larger number - some digits are pronounced differently when reciting the individual digits of a number, like in a phone number - phrasing like "one hundred 2" means 120, not 102 - large numbers not grouped in thousands, but in ten thousands, ie having a single word for 10000, a word for 10^7, etc, so a valid spoken number is "13 thousand ten-thousand" -> 1.3 billion

Image resolutions are also pronounced as 19-20 or 10-80 and not the long way.

These are American (and Canadian?) rules.

The British rules that differ are:

> As an exception to the above, a year like 2001, is spoken as two thousand and one, not as 20-01.

The vast majority of the time, but 20-01 is certainly understood, and is becoming more common as we progress through the century. (20-09 is more common than 20-01.)

Possibly because 2001 has an extra word in British English: "two thousand and one" rather than "two thousand one".

> Times like 3:45 are spoken in numerous ways, like 3-45, quarter to 4, 15-45 if afternoon, or possibly even 15 to 4.

Fifteen minutes to four. "Minutes" is usually excluded for 5, 10, 20 and 25, and nothing else.

But, there's usually a preferred choice of which way the time should be spoken, e.g. I can finish work at quarter past five, and take the seventeen thirty train home, which arrives at seventeen fifty-two, I'll be home by six.

> When giving an address, 386 Maple is said as 3-86 Maple, not as 3-8-6 nor the long the way.

So American! More than any of the others, should one want to sound British, don't say this.

In Britain, it's always either "three hundred and eighty six, Maple Street" or "three, eight, six, Maple Street".

> When speaking a series of digits, 0 can be said as oh or zero; in many contexts, oh is much more common.

Also nought, as in 0.003005, "nought point nought nought three oh oh five" (yes, really).

"Nil" for football, "we lost three - nil".

(And the dates, prices and measurements are all slightly different. Briefly, it's "the twenty-eighth of February" or [less common] "February the twenty-eighth", the "of" or "the" are essential. "One pound thirty", never like "a dollar thirty", and "fifty" pence/pee never "half". Either metric measurements, or a preference for different units if using traditional measurements -- stone over pounds for body weight, yards over feet for distances.)

Author here. I'm working on launching a new webapp dedicated to helping language learners master numbers both in production and listening comprehension. I'd love any feedback, tips, or suggestions from the HN community.

Can you add the names of the languages below the flags (probably both in English, and in that language as well)? It's not super easy to scan for the language I'm interested in based on flags alone, assuming I even know what the flag looks like.

Very nice idea.

On the main page, consider listing the languages, rather than using flags. Languages and countries don't always correspond one-to-one; many of the flags you have link to the same language, and conversely there are multiple languages spoken in some countries. A language list would be both more compact and more directly convey what people want to look for, rather than making people indirect through "I want to learn more of language X, which of the several countries speaking that language should I select, and does it make a difference?".

(That said, if you later expand to cover regional differences in accent, distinguishing between countries seems useful for that.)

Thanks! Regional dialects and differences in accent would be a great idea. I may end up being limited by the availability of the audio since I'm using a speech synthesis library at the moment. Perhaps some fallback methods of crowd-sourced audio for the less-common dialects would be an option.

This looks like it will be quite useful for anyone practicing languages that have their own word for 10,000 (Chinese, Japanese, etc), thanks!

In Chinese, the number "100,000" is not read as "100 thousand", rather it is "10 ten-thousands". This makes translating numbers in the range of 100,000-10,000,000 a bit awkward if you are not used to it.

Norwegian has a similar thing, mostly used by older folks.

"72" might be "seventy-two" but it is also "two plus seventy."

Hindi and related languages have the same thing. Lakh is 100k. Crore is 100 lakh.

This is really cool! When I was learning numbers in another language, I would say license plate numbers out loud (or in my head) each time I saw one. This would have been nice at that time!

As a side note, 'Finnish' is misspelled.

I tried the languages I know. Pretty cool! For my next language I'm starting now, this makes me want to start reviewing numbers even though it's not something I generally focus on at first.

Suggestion: make the lowest time at least 20 or 30 seconds, as 10 is too short.

Second suggestion: a visual component where you could utilize shop signs with prices on them (http://i.imgur.com/wsg5BU4.jpg) but I'm not sure how that could benefit the user, maybe the user has to type out the number (like "fifty seven").

Great app! Just a nitpick. Seems like China, Hong Kong and Taiwan all have "Chinese" selected. Where the dropdown menu in the freeplay page does have "Chinese Hong Kong" (or is it Cantonese?) provided.

I love German language, but one thing totally freaks me out is that in German numbers between 21 and 99 are pronounced "backwards". That is, 21 is "einundzwanzig" - "one and twenty". I speak German for somewhat 15 years already, but every time I have to deal with numbers below 100, I freeze for a couple of seconds.

same here

in german: when (for whatever reason) people dictate numbers i always insist on single digit dictation

my brain cant handle long series of "twoandtwenty, threeandforty"

Try French, where you eventually have to start doing addition, multiplication, or both...

Quatre-vingt-douze = four times twenty plus twelve = 92

Yes, but belgium and switzerland doesn't:

    France, belgium/switzerland
    70 -> soixante dix (60 + 10) -> septant (latin word like september for 7)
    80 -> quate vingt (4 * 20) -> octant/huitant (come from huit = 8)
    90 -> quatre vingt dix (4 * 20 + 10) -> nonant  (latin word for 9)

It's the same in Danish. And it repeats (in German too) for every 100, so you get "einhunderteinundzwanzig" in German or "(et) hundrede enogtyve" in Danish.

I imagine it can really trip you up with large numbers that are usually grouped by thousands, millions, billion, etc. when saying them out loud, like 123456789: "hundrede og treogtyve millioner, fire hundrede seksoghalvtreds tusinde og syv hundrede niogfirs", literally: "one hundred and three-and-twenty million, four hundred six-and-fifty thousand and seven hundred nine-and-eighty".

And Danish has yet another quirk. The names for 50, 60, 70, 80 and 90 aren' based on multiplication with 10 but on multiplication with 20, so instead of saying something to the effect of "five-tens" we say "half-third (i.e. 2.5) times 20", and then we even go and shorten it to obscure the fact that it's based on 20-multiplication.

I envy no one who has to learn Danish numerals.

English (and Swedish and Norwegian) all used to use the backwards form. (I know the English form mostly from an old nursery rhyme, with the line "four and twenty blackbirds" [1].) Before that, counting was also in twenties (score), as in "three score and ten".

I wish Danish had made the same changes. "Tres", "firs" are clearly something about three and four, and "halv" is half -- it's difficult not to think about 3, 4 and ½, and with the "sinds-tyve" omitted, there's no time to think before then next word is spoken. I also have difficulty distinguishing "halv" and "og" in "tre half tres" / "tre og tres".

This website will be useful practise.

I've found [2]

"A Scandinavist language reform movement tried to get the 20-based forms replaced by 10-based like Norwegian and Swedish have. With absolutely no success.

Danish 10-based forms are only used in inter-Scandinavian communication and money documents like cheques. They are: femti, seksti, syvti, ot(te)ti, niti"

I suspect English numbers would be better understood by the average Dane...

[1] https://en.wikipedia.org/wiki/Sing_a_Song_of_Sixpence#Lyrics

[2] http://www.olestig.dk/dansk/numbers.html

Counting in Japanese is easy to learn, because it is very consistent. For example, in English, we have "eleven," but in Japanese, it is "juuichi," or "ten one." "Twenty" is "nijuuichi," or "two ten one."

The thing that always throws me off is that the breaks in larger numbers are different than in English. For example, as we add more and more zeros to a number, we say:

One, ten, hundred, thousand, ten thousand, one hundred thousand, one million, ten million, one hundred million, one billion.

In Japanese, they break the numbers up like this (they use different words, of course):

One (1), ten (10), hundred (100), thousand (1,000), one "man" (10,000), ten "man" (100,000), one hundred "man" (1,000,000), one thousand "man" (10,000,000), one "oku," (100,000,000), and ten "oku" (1,000,000,000).

I'm pretty good up until one "man," then I have to start thinking.

But Japanese has the extra wrinkle that it uses different words depending on what is being counted...

That seemed insane to me at first but somehow that is easily intuitively absorbed.

The man (10,000) thing catches me out every time though.

Depending on how obscure the counter is - even Japanese people struggle [0]! Honestly, you only need a dozen or two counters. The hard part for me is remembering to use them.

[0] https://www.youtube.com/watch?v=N6sNa70KsW8

Not related to Japanese, but to large numbers in general: I just learned that there are to scales for billions - https://en.wikipedia.org/wiki/Long_and_short_scales

I'd never thought about it before, but Gujarati has the same issue. The word for 21 is pronounced like one-twenty.

I'm German and I often notice fellow Germans (including myself) getting two-digit numbers wrong when under stress, when translating things, or when dealing with lots of digits at the same time (phone numbers...). This is not a cute quirk of our language, but a wart that needs to go.

Interestingly, Czech allows reading numbers below 100 in this backward fashion, so 21 can be read as either "dvacet jedna" or "jednadvacet". I perceive the latter as less formal, but in everyday speech, nobody would raise an eyebrow if I used it. Likewise with ordinals: 21st could be pronounced either "dvacátý první" or "jednadvacátý".

One trick that helped me to master the numbers in a foreign languages was to mentally make additions in that language. For example, in Japanese, I would ask myself "How much is Ni plus Go?"

At first I would mentally do "well, Ni is 2, Go is 5, 2+5=7, 7 is Nana, so that's the answer" and at some point I was able to do it directly without translating the numbers to my native language.

That's a great idea for an advanced mode. Everyone loves math problems. :)

My Mandarin teacher did this. The answer was usually 14, 41, 40, 44 or similar. This was because, well:

  四是四,      sì shì sì
  十是十。      shí shì shí
  十四是十四,   shí sì shì shí sì
  四十是四十。  sì shí shì sì shí

4 is 4, 10 is 10, 14 is 14, 40 is 40

https://www.youtube.com/watch?v=5TuVL3mlBR4

The most difficult problem I have with Japanese numbers (going on 13 years of study and ten years in-country) is STILL dealing with counting higher numbers based on 10,000s. I've drilled and practiced and it still doesn't come fluently; I always find myself converting the Japanese to English, doing the math, then reconverting to Japanese, usually using my fingers to tick off zeroes to figure out how many 10,000s the number has.

The times I count higher than 1,000 in Japanese is almost always currency. I think in USD and yen, so if I say 一万円 I think either "ichimanen" or $100. When I have to translate a big number like 54万円 I just tell them $5400. It's surely not as good as properly learning but it just takes so much longer when translating live.

Mine only shows "American" on Firefox, plenty of other languages on Chrome, both on Windows 10. Why the discrepancy?

I'm not sure exactly. It is currently using the browser's speech synthesis functionality. I think Chrome has some languages shipped with it while Firefox may rely solely on the OS. I will probably need to switch away from the browser to get a consistent experience.

FF on Fedora 25 showed lots of languages, FYI.

My Firefox on Windows shows no languages available, as does Edge on Windows, while Chrome on Windows shows a handful of languages available.

Edit to say: what a fantastic and useful little tool! It is small focused, and very useful.

This is very cool. And it has Korean. When I tried to do Korean in Rosetta Stone, numbers are where I dropped off. With that said, it would be nice if there were a learning mode for people who don't know the numbers to learn them through this.

I am planning to add this! If you are interested, be sure to get on the list to know when it's released.

Korean has two sets of numbers! (Il, e, sam... and hana, dul, set, ... ). When I worked for a Korean company, I heard the first more than the second, but I could never work out what the right context was for each set of numbers. I could usually pick out the numbers up to 20 or so, which was all I really needed in a work context.

Not all languages in the "Lots of Supported Languages" are actually supported; i.e. Greek, Turkish, Swedish...

I'm learning Swedish, so that was the first I tried, and it worked perfectly for me. (Firefox 51.01, MacOS X 10.11)

This!

I was pretty excited to see a Norwegian flag on there, as not only do I sometimes not understand numbers, but there are a few different ways to say larger numbers. Click on the link and it is unavailable. I was sorely disappointed.

Hm, I'll look into it. Right now it is using the built-in SpeechSynthesis library in the browser. You may try installing the language in the accessibility settings for your OS. (Could you try this and let me know?)

I've considered adding a fallback API like Google's Cloud API if the browser is unavailable, but I haven't done that yet.

If you didn't, be sure to get on the feedback/email list and I'll let you know when I do.

Worked on Firefox on a Mac, but not on Firefox on Android.

I get about 40% correct in Norwegian, even though I've never studied the language before, since I'm studying Danish. I only get about 50% right in Danish :-(

"tre hundrede seks og halvfjerds"... mad! ("Three hundreds six and half-the-fourth-[multiple-of-twenty]", more-or-less.)

It was interesting when I was in Vietnam recently, a lot of Vietnamese people I dealt with outside the cities didn't speak any English at all, except for numbers, or at least pretended as much.

It seems a bit backwards really. It's easy to write down numbers (which they often did, or they showed me the cash I had to pay). Conveniently the motorbike mechanics are usually so good you just point at what's wrong and they'll figure it out pretty quickly.

Hmm everything I try is unavailable. Tried on firefox and chromium. Is the site overloaded or is it something on my end?

It's probably on your end as the voice is done in the browser. Which OS are you on? The browsers may require some OS functionality for the speech synthesis to work.

Slackware, firefox 51.0.1.

For Firefox on Linux you need to install a speech synthesis engine such as espeak

Same for me FF on Kubuntu, tried Chromium too.

Did work on Chrome - two problems: one was a sound cut off before the end and so I entered the wrong number; second the last number read finished the game but without any time for entering the number, it deducted a mark. A further game (different language) ended before I finished and didn't tell me the last answer.

Sometimes when I click Replay, it autofails me and moves to the next question.

This is great! Bookmarked. Numbers are rattled off so quickly in foreign languages, and small syntax differences where the order is changed and really cause new learners to pause and miss the complete rest of the sentence etc.

Just ran through Hungarian for a bit - lots of fun. Hungarian is easy as the numbering system is super consistent and logical. Though this just covers the cardinal numbers. The other types in Hungarian are quite different.

Really? Like what, for instance? I know that ordinal numbers are typically different in languages. Are there others?

Don't know about Hungarian, but in Russian I can think of at least four types of numerals:

- cardinal (what is the number?); for example, dva, tri, chetyre, pyat' (два, три, четыре, пять)

- ordinal (which position in a sequence?); for example, vtoroy, trety, chetvyorty, pyaty (второй, третий, четвёртый, пятый)

- collective (how many?); for example, dvoye, troye, chetvero, pyatero (двое, трое, четверо, пятеро)

- nominative (what is the number's name?); for example, dvoika, troika, chetvyorka, pyatyorka (двойка, тройка, четвёрка, пятёрка)

In Latin there are cardinals, ordinals, and also distributives, frequentatives, and according to some people multiplicatives (tres, tertius, terni, ter, triplex). Their meanings are something like:

three, third, three apiece, thrice, triple

These answer questions like: How many apples were there? Which one of the apples did you want? How were the apples distributed to the children? How many times did you count the apples? How has the apple population grown since last year?

At least for small numbers, English seems to have a distinctive form for each of these except for the distributive. But Latin will have many of these distinctive forms even for large numbers, like sescenties 'six hundred times' and octingentesimus 'eight hundredth'.

Edit: there's also some specialized vocabulary for fractions such as quadrans, sextans, octans, and even specific fractions like dodrans (3/4) and quincunx (5/12). But I'm not sure that these patterns are productive (generalizable) for larger denominators. Per Wikipedia, it looks like they're not, but quite a few have specific names:

https://en.wikipedia.org/wiki/Roman_numerals#Fractions

(I'd never heard of most of those! ... now I remember learning about the distributives for the first time at a spoken Latin event and remarking that numbers sounded surprisingly complicated in Latin, to which the organizer replied "dimidiam partem nescis!" 'you don't know the half of it!'.)

Cardinal, ordinal, fractional ordinal, numbers of people, Nth in a sequence and date in a month. This page explains it all with examples - http://www.hungarianreference.com/Numbers-cardinal-ordinal-s...

One day, during a multi-hour car-ride, I drilled my wife on French numbers by reading off the license plates we saw. She would translate two and three digits.

Her instructor said she had never seen someone improve so fast.

Nice one, bookmarked :) I'm trying it out now and one thing is a bit annoying - the track for the number is played immediately as I press the "guess" button for the previous one.

Yeah, it does. Do you think a short delay might be help?

This is amazing! Adding a "years" category would be great too!

Thanks! Great idea.

I entered 890 when correct answer was 89. 89 is all shown in green at the end and its hard to know which ones I got wrong. Other than that, really fun to play.

Btw, misspelt "Finnish" as "Finish" in multiple locations.

This is great! But I couldn't find any way to learn visually?

Yep, I plan to add this functionality soon.

All languages show as unavailable. Android 5.1.0 chrome 56

Cool, will be using it for Thai now. So far so good.

Your Australian voice is decidedly British.

The Australian voice is definitely Australian, and the British, British, for me.

I'm British. Are you British or Australian? I've found Americans often confuse us.

It's using browser speech synthesis rather than a recording, so the voice could be different from browser to browser and OS to OS!

Wow, this is actually really good. Thank you! This will help immensely for my Russian.

Any plans for ads or monetization, or are you just going to keep it as a free thing?

Reply on news.ycombinator.com