{
  "id": 5838,
  "title": "Marking",
  "url": "/competitions/facebook-recruiting-iii-keyword-extraction/discussion/5838",
  "author_name": "",
  "post_date": "2013-09-22T00:21:36.620Z",
  "votes": null,
  "comment_count": 1,
  "views": 1032,
  "content": "<p>Hi, I am very new to Machine Learning and Big Data and liked to start learning with this challenge as a starting point. My question is really basic, so please don't mind if this doesn't make any sense.</p>\n<p>I have taken following example from training dataset. Training dataset has only one tag &quot;css&quot; for following question but when I tried to predict tags for the same question then I came up with &quot;css, background-position, firefox and google-chrome&quot; as you can see in image.</p>\n<p>&nbsp;</p>\n<p><img src=\"http://oi40.tinypic.com/2mrtc09.jpg\" alt=\"question\" width=\"1599\" height=\"1064\"></p>\n<p>So my question is, these all predicted tags look relative to the question, how scoring system will score these predictions if we get this kind of question in test dataset and we predict like above predictions.</p>\n<p>&nbsp;</p>\n<p>Thanks</p>",
  "messages": [
    {
      "id": "31325",
      "postDate": "09/22/2013 00:21:36",
      "content": "<p>Hi, I am very new to Machine Learning and Big Data and liked to start learning with this challenge as a starting point. My question is really basic, so please don't mind if this doesn't make any sense.</p>\n<p>I have taken following example from training dataset. Training dataset has only one tag &quot;css&quot; for following question but when I tried to predict tags for the same question then I came up with &quot;css, background-position, firefox and google-chrome&quot; as you can see in image.</p>\n<p>&nbsp;</p>\n<p><img src=\"http://oi40.tinypic.com/2mrtc09.jpg\" alt=\"question\" width=\"1599\" height=\"1064\"></p>\n<p>So my question is, these all predicted tags look relative to the question, how scoring system will score these predictions if we get this kind of question in test dataset and we predict like above predictions.</p>\n<p>&nbsp;</p>\n<p>Thanks</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "31332",
      "postDate": "09/22/2013 08:38:26",
      "content": "<p>The predicted tags set has to be an exact match of the real tags set. So, for the prediction above you will get 1 true positive, 3 false positives and 0 false negatives.&nbsp;</p>\n<p>But this may be just an isolated case. In order to see if your algorithm has indeed too many false positives, you can compute an approximate F1 score offline, using only the training data.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 31332,
      "author_name": "andreibozantan",
      "author_url": "",
      "post_date": "09/22/2013 08:38:26",
      "content": "<p>The predicted tags set has to be an exact match of the real tags set. So, for the prediction above you will get 1 true positive, 3 false positives and 0 false negatives.&nbsp;</p>\n<p>But this may be just an isolated case. In order to see if your algorithm has indeed too many false positives, you can compute an approximate F1 score offline, using only the training data.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "31325": "",
    "31332": ""
  },
  "source": "meta"
}