{
  "id": 141499,
  "title": "Using all columns",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/141499",
  "author_name": "Suraj Parmar",
  "post_date": "2020-04-06T09:28:09.389000",
  "votes": 2,
  "comment_count": 5,
  "views": 0,
  "content": "<p>What about using other 5 labels combined with a NLP model embeddings?</p>",
  "messages": [
    {
      "id": 800728,
      "postDate": "2020-04-07T17:23:55.047Z",
      "content": "<p>That is a very good idea, which could have given a decent performance, however already tried that, it does give us good performance on local cv but almost same as only toxic on the leaderboard.</p>",
      "rawMarkdown": "That is a very good idea, which could have given a decent performance, however already tried that, it does give us good performance on local cv but almost same as only toxic on the leaderboard.",
      "votes": 2
    },
    {
      "id": 799231,
      "postDate": "2020-04-06T09:28:09.390Z",
      "content": "<p>What about using other 5 labels combined with a NLP model embeddings?</p>",
      "rawMarkdown": "What about using other 5 labels combined with a NLP model embeddings?",
      "votes": 2
    },
    {
      "id": 799759,
      "postDate": "2020-04-06T18:21:54.187Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 800072,
          "postDate": "2020-04-07T03:56:53.587Z",
          "content": "<p>Why wouldn't you?</p>",
          "rawMarkdown": "Why wouldn't you?",
          "votes": 1
        },
        {
          "id": 800094,
          "postDate": "2020-04-07T04:46:47.197Z",
          "rawMarkdown": "",
          "isDeleted": true
        },
        {
          "id": 800096,
          "postDate": "2020-04-07T04:49:12.477Z",
          "content": "<p>If you are new to kaggle, look into past discussions; These can be used easily for aux labels.</p>",
          "rawMarkdown": "If you are new to kaggle, look into past discussions; These can be used easily for aux labels.",
          "votes": 2
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 800728,
      "author_name": "Harshit Sheoran",
      "author_url": "",
      "post_date": "2020-04-07T17:23:55.047000",
      "content": "<p>That is a very good idea, which could have given a decent performance, however already tried that, it does give us good performance on local cv but almost same as only toxic on the leaderboard.</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 799759,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-04-06T18:21:54.187000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 800072,
          "author_name": "Aditya Soni",
          "author_url": "",
          "post_date": "2020-04-07T03:56:53.587000",
          "content": "<p>Why wouldn't you?</p>",
          "votes": 1,
          "replies": []
        },
        {
          "id": 800094,
          "author_name": "",
          "author_url": "",
          "post_date": "2020-04-07T04:46:47.197000",
          "content": "",
          "votes": 0,
          "replies": []
        },
        {
          "id": 800096,
          "author_name": "Aditya Soni",
          "author_url": "",
          "post_date": "2020-04-07T04:49:12.477000",
          "content": "<p>If you are new to kaggle, look into past discussions; These can be used easily for aux labels.</p>",
          "votes": 2,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "800728": "That is a very good idea, which could have given a decent performance, however already tried that, it does give us good performance on local cv but almost same as only toxic on the leaderboard.",
    "799231": "What about using other 5 labels combined with a NLP model embeddings?",
    "799759": ""
  }
}