{
  "id": 140249,
  "title": "Is there any advantages to use converted English translation?",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/140249",
  "author_name": "",
  "post_date": "2020-04-01T03:09:56.519761300Z",
  "votes": 6,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Just one more silly question, if my model gets trained on the translated eng data set, how about on the prediction phase? Do I need to convert non-eng to eng while prediction or not? Thanks.</p>",
  "messages": [
    {
      "id": "793440",
      "postDate": "04/01/2020 03:09:56",
      "content": "<p>Just one more silly question, if my model gets trained on the translated eng data set, how about on the prediction phase? Do I need to convert non-eng to eng while prediction or not? Thanks.</p>",
      "rawMarkdown": "Just one more silly question, if my model gets trained on the translated eng data set, how about on the prediction phase? Do I need to convert non-eng to eng while prediction or not? Thanks.",
      "votes": null
    },
    {
      "id": "793445",
      "postDate": "04/01/2020 03:13:40",
      "content": "<p>If your model is trained on translated english dataset, you would have to convert non-english data samples to english since your model only learns from english. It's just like if you were to teach a person to only know how to speak english, you cannot expect person to understand other languages like spanish. </p>\n\n<p>Also, in my opinion, it is not recommended to translate the english data since you are adding another level of uncertainty since this translation may not be 100% accurate.  <a href=\"/ipythonx\">@ipythonx</a> </p>",
      "rawMarkdown": "If your model is trained on translated english dataset, you would have to convert non-english data samples to english since your model only learns from english. It's just like if you were to teach a person to only know how to speak english, you cannot expect person to understand other languages like spanish. \n\nAlso, in my opinion, it is not recommended to translate the english data since you are adding another level of uncertainty since this translation may not be 100% accurate.  @ipythonx",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 793445,
      "author_name": "nicholasgah",
      "author_url": "",
      "post_date": "04/01/2020 03:13:40",
      "content": "<p>If your model is trained on translated english dataset, you would have to convert non-english data samples to english since your model only learns from english. It's just like if you were to teach a person to only know how to speak english, you cannot expect person to understand other languages like spanish. </p>\n\n<p>Also, in my opinion, it is not recommended to translate the english data since you are adding another level of uncertainty since this translation may not be 100% accurate.  <a href=\"/ipythonx\">@ipythonx</a> </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "793440": "Just one more silly question, if my model gets trained on the translated eng data set, how about on the prediction phase? Do I need to convert non-eng to eng while prediction or not? Thanks.",
    "793445": "If your model is trained on translated english dataset, you would have to convert non-english data samples to english since your model only learns from english. It's just like if you were to teach a person to only know how to speak english, you cannot expect person to understand other languages like spanish. \n\nAlso, in my opinion, it is not recommended to translate the english data since you are adding another level of uncertainty since this translation may not be 100% accurate.  @ipythonx"
  },
  "source": "meta"
}