in

Expanding the thought experiment – would you use a voice translator?

I started a new research process. Posting questions and getting responses to those questions online using Mylot. In part, I am doing that to get both sides and to hear arguments from both sides. I know that sometimes my world insulates me from hearing both sides. It is critical to hear and acknowledge the other side. Recently on MyLot, I posted an interesting question about voice translation hardware.  Some of the pushback from the other side is detailed./ Too complex, there are free products (Google Translate) and the most interesting argument the complexity of language makes it impossible for technology to solve the problem. Language, I am told is far too complex to build a machine to translate it. This is one area where humans remain superior.

(link to Mylot question and answers here)

First off, the question here is really what is the goal. I suspect the argument humans are better at this is based a little on fear and a little on a failure to distribute information. Based on that I would like to conduct a thought experiment. The thought experiment is detailed as follows. Fill a room with 30 people, all of whom do not speak the language of any of the other 30 people. Then, further, create separation by planning the Portuguese speaker and Spanish speaker at opposite ends of the room. Having learned Spanish as a child, I know Portuguese is close (not the same but close). Now give each of them a unique sentence to say and have a single person translate what they are saying. Have the sentence said once.

Now do the same thing with Google Translate. First of all, Google translates records what it is asked to translate. It can quickly break the structure of each sentence down and will convert the speech at a more effective rate in a much shorter period than the human would. The failure rate for the human will be higher than the machine, over a much longer period. Yes, the human would create perfect translations if they were given the paper each unique statement was written on and then given unlimited time to translate. The computer would do the same with the written scanned words, or the spoken. The computer would do so much faster.

All of this is simply easily proven now. Sadly he reality is that translation systems have improved in the past 12 months. The linear rate of improvement for the technology is significant.  The advantage of the fatter computer system is allowing it to share information with a human, that is relevant to what the person is seeking. Then the human can go and research the additional information required. The more complex a matching system is, the more highly tuned the Machine Learning system has to be. But, that is simply a process of time. We could do the same experiment with 30 speakers of English, all with different dialects. Each of the English speakers having not only regionally unique words but also dialectically unique words. A human might know 10, 12 of those unique meaning. The computer would know all 30. The power of technology is its ability to free humans from the mundane!

  • Question of

    Do you speak more than 20 languages?

    • Yes
    • No
  • Question of

    Do you use google translate?

    • Yes
    • No
  • Question of

    Have you ever been in a sitation where not knowing what was being said impacted you directly?

    • Yes
    • No

Report

What do you think?

Legend

Written by DocAndersen

One fan, One team and a long time dream Go Cubs!!!!!!!!!!!!!

12 Comments

Leave a Reply
  1. As a part time medical translator, I know for a fact that some translation software are better than others. For example SDL Trados is good along with Memsource and MemoQ. But there is one thing that cannot be matched by a translation software and that is sentence structure and most terminology. Now here I am talking just plain English to French translation and medical translation. Even though you can add terminology to the software dictionary, most of the time the sentence wording is still off. But I do agree that translation software have come a long way. BTW, I do not own a translation software but I will use Microsoft translate to get me a baseline from which I can start translating.

    1
    • Machine Learning (which allows the computer to do a better job of pattern matching) and the increasing speed of computers will solve the construction of sentence issue it just takes a little time.

      Now, you are talking however about a profession that I suspect will exist long after translation software takes over. Medical translation has a lot more “human” requirement than other translations!

      1
  2. I do second that machine cannot translate 100% and can cause jokes. I’ve tried to translate Chinese to English and vice versa … sometimes I get really bad and nonsensical translations. I’ve copied and pasted Japanese text from websites into Google Translate and the text is comprehensible. (but of course I don’t know Jap, so maybe the report is nonsensical by itself?)

    1
    • Disagree with the reality, although you have picked the one language that even human translators fail at translating.

      Jokes however are something that most humans fail to get from one language to another.

      I guess not sure I agree with either of your points. But do agree things need to continue to get better.

      1
      • Sorry, I may have not been clear myself by “joke” I don’t mean I translate jokes but the translation are “jokes” – like too funny to not laugh at it.

        While machines can translate most sentences well enough, there are much they cannot do yet. But of course I can only say for the 2 languages I know well enough to read and write. Maybe they’re much better in other languages.

        I do know such translators will eventually become better and maybe even take over human translator but I don’t see that in the near future.

        1

Leave a Reply