{
  "id": 349660,
  "title": "AI4Code with multiple languages translation",
  "url": "/competitions/AI4Code/discussion/349660",
  "author_name": "",
  "post_date": "2022-09-02T07:55:20.371818400Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<p>In this competition we had notebooks from Kaggle in many different languages. This could be handled with example multilingual models etc.<br>\nBut I also did I try with a code that automatically detect and translate any given language that came up in the hidden test data.</p>\n<p>In the below notebook I have shared the code but in short:<br>\nI used xlmrobertabaselanguagedetection for detection, used lang. with probability over 0.99 and excluded \"Eng\", \"unknown\" and \"It\".<br>\nFor the translation I used M2M100!</p>\n<p><a href=\"https://www.kaggle.com/code/kirderf/ai4code-with-languages-translation\" target=\"_blank\">https://www.kaggle.com/code/kirderf/ai4code-with-languages-translation</a></p>\n<p>All in all, it gave a score boost vs using the same NLP models without translation.</p>\n<p>That’s it!</p>",
  "messages": [
    {
      "id": "1923411",
      "postDate": "09/02/2022 07:55:20",
      "content": "<p>In this competition we had notebooks from Kaggle in many different languages. This could be handled with example multilingual models etc.<br>\nBut I also did I try with a code that automatically detect and translate any given language that came up in the hidden test data.</p>\n<p>In the below notebook I have shared the code but in short:<br>\nI used xlmrobertabaselanguagedetection for detection, used lang. with probability over 0.99 and excluded \"Eng\", \"unknown\" and \"It\".<br>\nFor the translation I used M2M100!</p>\n<p><a href=\"https://www.kaggle.com/code/kirderf/ai4code-with-languages-translation\" target=\"_blank\">https://www.kaggle.com/code/kirderf/ai4code-with-languages-translation</a></p>\n<p>All in all, it gave a score boost vs using the same NLP models without translation.</p>\n<p>That’s it!</p>",
      "rawMarkdown": "In this competition we had notebooks from Kaggle in many different languages. This could be handled with example multilingual models etc.\nBut I also did I try with a code that automatically detect and translate any given language that came up in the hidden test data.\n\nIn the below notebook I have shared the code but in short:\nI used xlmrobertabaselanguagedetection for detection, used lang. with probability over 0.99 and excluded \"Eng\", \"unknown\" and \"It\".\nFor the translation I used M2M100!\n\nhttps://www.kaggle.com/code/kirderf/ai4code-with-languages-translation\n\nAll in all, it gave a score boost vs using the same NLP models without translation.\n\nThat’s it!",
      "votes": null
    },
    {
      "id": "1923453",
      "postDate": "09/02/2022 08:36:27",
      "content": "<p>Thanks for sharing, great resource for further competions with multi langs!</p>",
      "rawMarkdown": "Thanks for sharing, great resource for further competions with multi langs!",
      "votes": null
    },
    {
      "id": "1923467",
      "postDate": "09/02/2022 08:49:18",
      "content": "<p>Thanks! Sharing is part of the community teamwork!</p>",
      "rawMarkdown": "Thanks! Sharing is part of the community teamwork!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1923453,
      "author_name": "goldenlock",
      "author_url": "",
      "post_date": "09/02/2022 08:36:27",
      "content": "<p>Thanks for sharing, great resource for further competions with multi langs!</p>",
      "votes": null,
      "replies": [
        {
          "id": 1923467,
          "author_name": "kirderf",
          "author_url": "",
          "post_date": "09/02/2022 08:49:18",
          "content": "<p>Thanks! Sharing is part of the community teamwork!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1923411": "In this competition we had notebooks from Kaggle in many different languages. This could be handled with example multilingual models etc.\nBut I also did I try with a code that automatically detect and translate any given language that came up in the hidden test data.\n\nIn the below notebook I have shared the code but in short:\nI used xlmrobertabaselanguagedetection for detection, used lang. with probability over 0.99 and excluded \"Eng\", \"unknown\" and \"It\".\nFor the translation I used M2M100!\n\nhttps://www.kaggle.com/code/kirderf/ai4code-with-languages-translation\n\nAll in all, it gave a score boost vs using the same NLP models without translation.\n\nThat’s it!",
    "1923453": "Thanks for sharing, great resource for further competions with multi langs!",
    "1923467": "Thanks! Sharing is part of the community teamwork!"
  },
  "source": "meta"
}