{
  "id": 75845,
  "title": "Data Augmentation",
  "url": "/competitions/quora-insincere-questions-classification/discussion/75845",
  "author_name": "",
  "post_date": "2018-12-27T04:57:07.856620Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Has anyone tried augmenting the data? It has been done before in another 'toxic comments' competition. See: <a href=\"https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py\">https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py</a></p>",
  "messages": [
    {
      "id": "445813",
      "postDate": "12/27/2018 04:57:07",
      "content": "<p>Has anyone tried augmenting the data? It has been done before in another 'toxic comments' competition. See: <a href=\"https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py\">https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py</a></p>",
      "rawMarkdown": "Has anyone tried augmenting the data? It has been done before in another 'toxic comments' competition. See: https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py",
      "votes": null
    },
    {
      "id": "446000",
      "postDate": "12/27/2018 10:33:50",
      "content": "<p>That kind of augmentation requires internet access (TextBlob's <code>translate()</code> feature uses the Google Translate API under the hood). For this competition, internet access is not allowed.</p>",
      "rawMarkdown": "That kind of augmentation requires internet access (TextBlob's `translate()` feature uses the Google Translate API under the hood). For this competition, internet access is not allowed.",
      "votes": null
    },
    {
      "id": "446147",
      "postDate": "12/27/2018 15:28:18",
      "content": "<p>\ndef data_argumentation(data, num=50000):\n    result = []\n    for i in range(num):\n        random.seed(i)\n        b = random.sample(data, 2)\n        result.append(b[0] + ' ' + b[1])\n    return np.array(result)\n\nI try this way to add positive data,but overfit = = </p>",
      "rawMarkdown": "<pre>def data_argumentation(data, num=50000):\n    result = []\n    for i in range(num):\n        random.seed(i)\n        b = random.sample(data, 2)\n        result.append(b[0] + ' ' + b[1])\n    return np.array(result)\n<code>\nI try this way to add positive data,but overfit = = </code></pre>",
      "votes": null
    },
    {
      "id": "446308",
      "postDate": "12/27/2018 21:02:42",
      "content": "<p><em>no external data</em> rule wont allow that :)</p>",
      "rawMarkdown": "_no external data_ rule wont allow that :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 446000,
      "author_name": "mschumacher",
      "author_url": "",
      "post_date": "12/27/2018 10:33:50",
      "content": "<p>That kind of augmentation requires internet access (TextBlob's <code>translate()</code> feature uses the Google Translate API under the hood). For this competition, internet access is not allowed.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 446147,
      "author_name": "mldevl",
      "author_url": "",
      "post_date": "12/27/2018 15:28:18",
      "content": "<p>\ndef data_argumentation(data, num=50000):\n    result = []\n    for i in range(num):\n        random.seed(i)\n        b = random.sample(data, 2)\n        result.append(b[0] + ' ' + b[1])\n    return np.array(result)\n\nI try this way to add positive data,but overfit = = </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 446308,
      "author_name": "abhishek",
      "author_url": "",
      "post_date": "12/27/2018 21:02:42",
      "content": "<p><em>no external data</em> rule wont allow that :)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "445813": "Has anyone tried augmenting the data? It has been done before in another 'toxic comments' competition. See: https://github.com/PavelOstyakov/toxic/blob/master/tools/extend_dataset.py",
    "446000": "That kind of augmentation requires internet access (TextBlob's `translate()` feature uses the Google Translate API under the hood). For this competition, internet access is not allowed.",
    "446147": "<pre>def data_argumentation(data, num=50000):\n    result = []\n    for i in range(num):\n        random.seed(i)\n        b = random.sample(data, 2)\n        result.append(b[0] + ' ' + b[1])\n    return np.array(result)\n<code>\nI try this way to add positive data,but overfit = = </code></pre>",
    "446308": "_no external data_ rule wont allow that :)"
  },
  "source": "meta"
}