View original (French)

Create a smarter translator than Google Translate in Java

dlife123 Posted messages 44 Status Member -  
 SpeezZealot -
Hello,
I'm looking to create a translator that's smarter than Google Translate.
But I can't find an open source dictionary to find words, their meanings, and their equivalents in English.
I know that if Google Translate isn't perfect, it means the task isn't easy, but I would still like to get started.
If anyone has a lead, I'm interested.
Thank you in advance.

4 answers

  2.
    dlife123 Posted messages 44 Status Member 35
     
    thank you for your response
    I followed the video and it doesn't contradict what I said before: "the task is not easy"
    but it must change, I will probably fail but I want to try
    all I'm asking is where to find the words and their meanings
    0
  3. hopelan Posted messages 114 Status Member 84
     
    Hello,
    If you still want to try, you would need to code search bots and equip yourself with storage servers in order to index a number of multilingual texts as well as texts containing expressions specific to each language. Then, for sorting, you just need to develop algorithms based on the referencing of these texts to assess their credibility. Finally, language dictionaries, expressions, and semantics will allow you to refine your translation proposals.

    Not being an expert, I'm simply making suggestions based on my knowledge in this area. Some members of this forum will certainly have things to say since the complexity level of a translator like Google is really very high. For the algorithms, you will need to refer to research articles to develop your own, but those of Google remain secret.
    0
  4. SpeedZealot
     
    Hello,

    For the concept, I think a language like Prolog will be easier (even taking into account the learning time of the language)

    For the dictionary, you can always create a program (Java for this) that inputs each French word from the list http://www.listesdemots.fr/touslesmots.txt into Google Translate and reads its translation

    However, I don't know how to do that, but it should be feasible.
    0
    1. dlife123
       
      Thank you very much for your response and the list of words.
      I would now like to classify these words according to their nature: verbs, articles, etc., which is very simple.
      I would also like to classify the verbs according to whether they apply to humans, animals, birds, etc.
      My program needs to be a little bit intelligent.
      For example, it should know that a human eats but not a house.
      We build a road but not a beast.
      I hope I was clear.
      Thank you very much for your help.
      0
      1. madmyke Posted messages 52304 Registration date   Status Moderator Last intervention   12 486 > dlife123
         
        Uh...

        "I hope I was clear. Thank you very much for your help."

        Why do you think linguists and programmers have been struggling with this type of program for 20 years? And it's not for lack of trying.

        Why do you think even organizations like the EU, with huge resources, still use thousands of translators to review automatic translations? (I worked in that department for 6 months in 2014)

        Let's not even talk about Google, whose immense resources face the same difficulties.

        Because it doesn't boil down to a person waking up in the morning with the idea that "the program must be a tiny bit intelligent." There are people who have been doing this for a long time, outstanding computer scientists aided by seasoned linguists. It has progressed over 20 years, but the problem remains the same.

        Translation is not (by a long shot) just about an excellent dictionary and some logical rules, even complex ones. Language is already extraordinarily complicated in itself, as it is impossible for a computer, even a powerful one, to differentiate a "context" (which you have already highlighted); grammar can be partially translated into algorithms, but there are tons of contextual things that, for now, only your brain can understand. We're not talking about punctuation (Tonight we eat the children <=> Tonight we eat, the children) and so on.
        If you have to add another level of translation, it becomes frankly unfeasible.
        If it was just some brilliant programming, it would have been done a long time ago.

        I know I might be harsh, but be realistic: language (outside of computer science) is an extremely complicated thing. Unless you are the new Einstein in programming and (especially) in language structure and everything related to it, I doubt you'll get anywhere other than just lining up words and translating them word for word. But there is already a plethora of software that does that and does it even better.

        Best regards.
        0
      2. SpeezZealot > madmyke Posted messages 52304 Registration date   Status Moderator Last intervention  
         
        Let him experiment! He has the right to try after all.

        For the classification of verbs, it's quite feasible with the dictionary: you can differentiate verbs from others by their conjugation (which is found in the list).

        As for the issue of a dog that cannot be constructed, the only idea I have is to use Google Images: after searching for "construct dog," if you find something other than a dog, it means it’s not applicable. With that, you can make classifications (e.g., the non-constructibles) which will be classified internally as "animal," "vegetable," etc. And so on.

        Well, I don't know if it's feasible, but it’s an idea ^^
        0