from Hacker News

The Shallowness of Google Translate (by Douglas Hofstadter)

by jsomers on 2/2/18, 12:11 AM with 24 comments

  • by skybrian on 2/2/18, 2:36 AM

  • by paulsutter on 2/2/18, 2:10 AM

    Superb reminder of the difficulty of true NLP. Google Translate is an amazing and very useful tool, but machines have a long way to go before they really understand language.

    Example sentence used by article:

    > In their house, everything comes in pairs. There’s his car and her car, his towels and her towels, and his library and hers

  • by oh-kumudo on 2/2/18, 2:15 AM

    Seems Google has update their translation model?

    For the Chinese one, at least:

    >> After one year of working in Tsinghua University, Zhong Shu was transferred to Mao's translation committee to live in the city and back to school on weekends. He still holds the post of graduate student.

    The leader of the Mao Selected Translation Committee is Comrade Xu Yonglian. Introducing Zhong Shu to do this job is Tsinghua classmate Qiao Guanghua.

    On the appointed day, after dinner, an old friend hired a rickshaw to come from the city to congratulate. After the guests go, Zhong book said to me in fear:

    He thought I had to do a "Southern study walk." This is not a good thing to do.

    >>

    Now they correctly singles out person's name, "锺书(Zhong Shu)", as comparing to transliterate it as Book(meaning of the character Shu). Even with that 南书房行走，IMO, it did a not bad job, at least knowing it is its own entity, not to break into parts then translate.

    As a native speaker, the style of the example text provided is quite elegant and old-school. 南书房行走 is a very confusing phrase, it looks like a verb but used as a noun phrase, and without context, it is hard for me to tell the meaning of it.

    The updated version of the translation is pretty serviceable. Google Translate works best with functional text, like news/report, etc. Not quite there with literature, which is well known, probably on purpose. As someone works on MT project, this quality is pretty amazing. I won't necessarily say it is shallow, TBH.

  • by gwern on 2/2/18, 2:19 AM

    I wonder about those samples. The Chinese I would expect to be kind of garbage going off BLEUs from published papers, but the French and German should be great. Am I imagining it, or do they just not sound like RNN translations? The RNN translations in my experience tend to flow well and are comprehensible, albeit sometimes just wrong because they picked a wrong semantic interpretation. But these samples sound like the older phrase-based translations, where they immediately descend into word salad. As people noted at the time, the jump in quality for French/German was so large that people knew when the new translations went live without Google publishing anything... (I assume Hofstadter at least was careful enough to do his sampling after the new RNN translations were opened up.)

    Google doesn't discuss how it rolls out the translation upgrades, exactly, and it's an uneven deployment. Can any Googlers comment on the possibility Hofstadter was using the old translations? Or can any NMT researchers compare and contrast his examples with current SOTA models?

  • by CurtHagenlocher on 2/2/18, 1:44 AM

    Hofstadter's "Le Ton beau de Marot" (which is about translation) is one of my favorite books of all time.

  • by ceautery on 2/2/18, 1:36 AM

    Hofstadter is still a boss. I felt like I was reading Metamagical Themas again.

  • by cooper12 on 2/2/18, 2:50 AM

    Great article that tackles some of the unwarranted hype machine learning has gotten lately, with everyone foaming at the mouth about Skynet and their jobs becoming obsolete. The author also tackles the common excuse of "it'll only get better". One interesting highlight of the article for me was the parenthetical around his translation of a paragraph in Chinese: "it took me hours". Translation is so much more than vectors between words in two languages and this article expresses that quite clearly.

  • by ronilan on 2/2/18, 2:50 AM

    Google translate, like a screwdriver is a tool. One can possibly do nasty things with it, but one, if is not a tool, and if limits oneself to just screwing things, should be ok.

    Can we translate that? Naaa. But that doesn’t make the tool shallow.

    P.S - I use the following trick to improve the odds of a good translation - the phrase has to be a “stable Google translate triangulation”.

    This is when a phrase does not change while switching back and fourth between three languages, two of which you know at a native level.

    Dope for me. YMMV. :)

  • by fenomas on 2/2/18, 2:51 AM

    Always nice to read Hofstadter, but in my experience Google Translate often returns gibberish even for single words, never mind subtleties of grammar. I don't think it replaces a foreign language dictionary yet, so arguments that it doesn't replace human translators seem like overkill!

    Random examples for Japanese: for "七輪" (brazier) it returns "tambourine", for "ちゃぶ台" (tea table) it returns "Shabu-bashi".

  • by mirimir on 2/2/18, 3:07 AM

    If you don't know any Chinese, and you need to use a website in Chinese, what's the alternative?

    I like the Firefox add-on Perapera for Chinese and Japanese. You hover, and see pop-up translation.

    Edit: style

  • by hyperpallium on 2/2/18, 3:04 AM

    I tell to you, although he appears as the friendly interlocutor, this fellow is an interrogator of the utmost cunning, with the most sinister goal... of entrapment!

  • by debt on 2/2/18, 2:26 AM

    Rev works really well but an actual human that transcribes what you say and it takes a day.