OSDev.org

And then there's stuff like this, which is why achieving human-level machine translation will require considerable contextual disambiguation using a semantic model of the world:

Oh yes, this can also happen in Estonian (although this particular example has different Estonian translations - spruce = kuusk, six = kuus, moon = kuu). For example, kuu = moon / month, tee = tea / road. I once heard the word "jäätee" on the radio and immediately thought about ice tea (which is a perfectly valid translation), but from the context it became clear that they were talking about the ice road over the frozen Baltic Sea, which is usually open in deep winter.

Nice one.

Though having words with multiple meanings isn't something specific to Finnish or Estonian. Let's try this:

Den Läufer umfahren - to drive around the runner
Den Läufer umfahren - to run over the runner
Den Läufer umfahren - to drive around the young pig
Den Läufer umfahren - to run over the young pig
Den Läufer umfahren - to drive around the carpet
(Okay, knocking down a carpet doesn't make sense, so here only one meaning of "umfahren".)
Den Läufer umfahren - to drive around the bishop (chessman)
Den Läufer umfahren - to run over the bishop (chessman)
...
(Other meanings of "Läufer" in Wikipedia)

This is also possible in Russian, where the meaning depends on the word's pronunciation (as in your example umfahren vs. umfahren):

дома - houses, дома - at home
он стоит - he / it stands, он стоит - he / it costs

XenOS wrote:For example, kuu = moon / month, tee = tea / road. I once heard the word "jäätee" on the radio and immediately thought about ice tea (which is a perfectly valid translation), but from the context it became clear that they were talking about the ice road over the frozen Baltic Sea, which is usually open in deep winter.

In Finnish, "jäätee" would be the ice tea, "jäätie" the ice road. Just generally different enough to make these two distinct languages instead of dialects. (This case being a more subtle difference than on average.)

Kevin wrote:Though having words with multiple meanings isn't something specific to Finnish or Estonian. Let's try this:

Den Läufer umfahren - to drive around the runner
[...]
(Other meanings of "Läufer" in Wikipedia)

Nice example, bookmarked it. Is that a specific, often-quoted ambiguity, or are there tons more that could be constructed?

"umfahren" is probably the example to show that stress can matter in German, though I generally know it as "den Polizisten umfahren" (to drive around/knock down the policeman). "Läufer" was just what I spontaneously could find that combines well with "umfahren" because many of the meanings are actual phyiscal objects that can be knocked down. You can probably find many more, and if you also include abstract concepts, definitely tons of them.

Arto wrote:In Finnish, "jäätee" would be the ice tea, "jäätie" the ice road. Just generally different enough to make these two distinct languages instead of dialects. (This case being a more subtle difference than on average.)

Although the difference between language and dialect is often based on whether there exists a change in grammar, rather than just words or pronounciation.

But even more often it's based purely on politics. As the famous quote says:

A shprakh iz a dialekt mit an armey un flot.
("A language is a dialect with an army and navy.")

Counterexample: There is no language called "American". It's just English be it with added typos like we expect from any dialect.

Counterexamples to statements qualified with "more often" don't actually add much.

Combuster wrote:Counterexample: There is no language called "American". It's just English be it with added typos like we expect from any dialect.

On that front, I saw this amusing bit the other day:

Fair description, I guess.

Arto wrote:

Combuster wrote:Counterexample: There is no language called "American". It's just English be it with added typos like we expect from any dialect.

On that front, I saw this amusing bit the other day:

OK, that just makes me laugh. BTW: this is a US-based company: http://www.valvesoftware.com/contact/ (or at least what my ISP allows me to see).

That "screenshot" has been photoshopped, it appeared on imgur a few days ago. The real language descriptions are "English (UK)" and "English (US)", IIRC.

Maybe they should add English (Australia)