So I was poking around Google Translate recently, and I noticed that they added an option for Hungarian. Hungarian being the extremely regular and extremely simple to parse language that it is, I was always surprised why nobody has made a machine translator for it (apart from this one really awful one that tried to treat the noun case propositions as propositions, leading to such nonsense as "ban a szobá". But that didn't do any grammatical analysis at all (it was just direct translation) and it just translated words directly and seemed like a halfhearted effort at just making a translator for every single language that had a dictionary on that service, so I didn't think much of it.)
Anyway, I decided to test out just how good it was. Decided to start with HU->EN, and for my test I was using the first couple paragraphs of Egri Csillagok, a fantastic Hungarian novel. So, let's see what we get:
The only thing notable about the first paragraph is that it translated "fürdik" correctly but left "fürödnek" untranslated, which immediately shows they're storing separate forms for each verb in a language with like 12 irregular verbs, 9 of which work in the same way and then only 3 (van, jön, megy) that are truly irregular.
Anyway, things proceeded to get interesting:
"Az erdőben jártak, patakra találtak." -> "The forest had, found a stream."
I'm not even sure how it got that. It completely ignored the postposition marking "in" and then the verb itsself somehow! It's as if the translation app crashed mid-sentence or something.
"Először csak a lábukat mártogatták bele, azután beleereszkedtek térdig." -> "First, only the feet mártogatták it, then beleereszkedtek knees."
Which is another charming example of Google Translate not translating what are basically extremely regular verbs. More perplexingly, it will translate "mártogat" correctly (as "to sop") but it will not translate any conjugation of it whatsoever.
I did a little more poking around but most of the errors were either pertaining to verb conjugation or really bizarre stuff like the one in "Az erdőben jártak". Then I decided to see how EN->HU was. The results were...hilarious and perplexing:
First, my usual trial sentence, "I am working on assignments" (which in Google Translate Polish gives you the lovely translation of "Jestem pracy na zadania".), which translated as:
"Én dolgozó megbízásokat."
Great, so it keeps trying to use the participle form instead of uh you know conjugating the verb in the first person singular.
Testing out some stuff with posessive suffixes, we get the wonderful pair of:
"My book" -> "könyvem" (correct.)
"My books" -> "Saját könyvek" (uhh...) (FWIW, it should be könyveim for whoever's wondering.)
"I read my book" and "I read your book" BOTH translate as "Elolvastam a könyvet".
"I have a computer" -> "Van egy számítógép" (now apparently changed to "Nekem van egy számítógépre", but "You have a computer" is still "Van egy számítógép". Also if you're using Nekem, isn't it the case that the noun still needs the posessive case (-em in this case) and not -re?)
Translating "Van egy számítógép" back we get "Do you have a computer?", which, uh, what?
And finally, this last bit, which is seriously the most brain-breaking thing that I have EVER seen come out of a machine translator:
"I am reading books" -> "Én olvasni" (I to read)
"I am reading a book" -> "Én olvasott" (I he/she/it has read)
"I am reading the book" -> "Én mint a könyv" (I like/as the book)
Now, curiously, if you use "I read..." instead, it translates it with olvas in the past tense, but it actually does it correctly. Curiously though it uses olvastam with *a* book and elolvastam with *the* book, but this is not as egregious as the 3 lines above which are simply...uh, yeah.
I would have tried seeing how well it handled translating into things like Hungarian definite vs. indefinite verbs in certain sentences as well as trying to see how well it would handle verbs with prefixes (it likely stores a table of verbs with a prefix for each separate verb) but until they start getting like the most basic stuff correct I'm not even going to bother, it'd be like shooting fish in a barrel.
Finally, I tried to see if Polish->Hungarian might give something that would be remotely comprehensible, on the principle of "2 shitty things may result in a good one". Instead I had the sentence "Jestem Google Translate i jestem bardzo ohydny i niedobry" ("I'm google translate and I'm very horrible and bad") turn into "Google Translate Én és én vagyok nagyon szerencsétlen, és rossz." ("Google Translate I and I am very unlucky, and bad."), which...yeah. (Also it managed to translate "ohydny" as "névtelen", "szerencsétlen" and "boldogtalan" just by mildly permuting the sentence, i.e. "Ja jestem..." used "boldogtalan" instead of "szerencsétlen").
Man, at this point I'm not sure what to do. Probably email google, and also maybe try working on my own attempt at an HU/EN translator? I've been putting it off for far too long and I might finally have enough motivation now.