{
  "id": 243405,
  "title": "What are we predicting?",
  "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/243405",
  "author_name": "",
  "post_date": "2021-06-02T12:22:59.608992400Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Honestly we are so late in the competition but I haven't seen this stated clearly anywhere. Are we predicting <em>ALL</em> datasets mentioned in a particular paper or the dataset/datasets <em>USED</em> by the author? </p>\n<p>So if I extract an actual dataset <strong>mentioned in the paper</strong> , can it be false positive or it will always be a true positive as long as it is, indeed a dataset? </p>\n<p>Thanks in advance</p>",
  "messages": [
    {
      "id": "1332998",
      "postDate": "06/02/2021 12:22:59",
      "content": "<p>Honestly we are so late in the competition but I haven't seen this stated clearly anywhere. Are we predicting <em>ALL</em> datasets mentioned in a particular paper or the dataset/datasets <em>USED</em> by the author? </p>\n<p>So if I extract an actual dataset <strong>mentioned in the paper</strong> , can it be false positive or it will always be a true positive as long as it is, indeed a dataset? </p>\n<p>Thanks in advance</p>",
      "rawMarkdown": "Honestly we are so late in the competition but I haven't seen this stated clearly anywhere. Are we predicting *ALL* datasets mentioned in a particular paper or the dataset/datasets *USED* by the author? \n\nSo if I extract an actual dataset **mentioned in the paper** , can it be false positive or it will always be a true positive as long as it is, indeed a dataset? \n\nThanks in advance",
      "votes": null
    },
    {
      "id": "1333140",
      "postDate": "06/02/2021 14:02:43",
      "content": "<p>Hi, this is from the evaluation page: <br>\n<em>The objective of the competition is to identify the mention of datasets within scientific publications. Your predictions will be short excerpts from the publications that appear to note a dataset.</em></p>\n<p>So yes, anything 'that appears noting a dataset',, go for it.<br>\nThe key is that you have to include 'each and every mention'  evenif they could be for the same dataset, as long as the wording of that specific mention is different from your existing predictions.</p>\n<p>There is a potential exception though: There have been early discussions on cases in public sets (train/test) where one mention seems to be producing more than one Ground Truth (a longer version of the mention and a shorter one). I haven't seen the issue resolved, so I reraised it before the final weeks but, hey good luck to all of us who participate 😃. </p>",
      "rawMarkdown": "Hi, this is from the evaluation page: \n*The objective of the competition is to identify the mention of datasets within scientific publications. Your predictions will be short excerpts from the publications that appear to note a dataset.*\n\nSo yes, anything 'that appears noting a dataset',, go for it.\nThe key is that you have to include 'each and every mention'  evenif they could be for the same dataset, as long as the wording of that specific mention is different from your existing predictions.\n\nThere is a potential exception though: There have been early discussions on cases in public sets (train/test) where one mention seems to be producing more than one Ground Truth (a longer version of the mention and a shorter one). I haven't seen the issue resolved, so I reraised it before the final weeks but, hey good luck to all of us who participate 😃.",
      "votes": null
    },
    {
      "id": "1333438",
      "postDate": "06/02/2021 17:58:45",
      "content": "<p>Good luck, we've got this!</p>",
      "rawMarkdown": "Good luck, we've got this!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1333140,
      "author_name": "abdulkadirguner",
      "author_url": "",
      "post_date": "06/02/2021 14:02:43",
      "content": "<p>Hi, this is from the evaluation page: <br>\n<em>The objective of the competition is to identify the mention of datasets within scientific publications. Your predictions will be short excerpts from the publications that appear to note a dataset.</em></p>\n<p>So yes, anything 'that appears noting a dataset',, go for it.<br>\nThe key is that you have to include 'each and every mention'  evenif they could be for the same dataset, as long as the wording of that specific mention is different from your existing predictions.</p>\n<p>There is a potential exception though: There have been early discussions on cases in public sets (train/test) where one mention seems to be producing more than one Ground Truth (a longer version of the mention and a shorter one). I haven't seen the issue resolved, so I reraised it before the final weeks but, hey good luck to all of us who participate 😃. </p>",
      "votes": null,
      "replies": [
        {
          "id": 1333438,
          "author_name": "giorgosmaragkopoulos",
          "author_url": "",
          "post_date": "06/02/2021 17:58:45",
          "content": "<p>Good luck, we've got this!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1332998": "Honestly we are so late in the competition but I haven't seen this stated clearly anywhere. Are we predicting *ALL* datasets mentioned in a particular paper or the dataset/datasets *USED* by the author? \n\nSo if I extract an actual dataset **mentioned in the paper** , can it be false positive or it will always be a true positive as long as it is, indeed a dataset? \n\nThanks in advance",
    "1333140": "Hi, this is from the evaluation page: \n*The objective of the competition is to identify the mention of datasets within scientific publications. Your predictions will be short excerpts from the publications that appear to note a dataset.*\n\nSo yes, anything 'that appears noting a dataset',, go for it.\nThe key is that you have to include 'each and every mention'  evenif they could be for the same dataset, as long as the wording of that specific mention is different from your existing predictions.\n\nThere is a potential exception though: There have been early discussions on cases in public sets (train/test) where one mention seems to be producing more than one Ground Truth (a longer version of the mention and a shorter one). I haven't seen the issue resolved, so I reraised it before the final weeks but, hey good luck to all of us who participate 😃.",
    "1333438": "Good luck, we've got this!"
  },
  "source": "meta"
}