{
  "id": 228324,
  "title": "External Data so early",
  "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/228324",
  "author_name": "",
  "post_date": "2021-03-24T06:54:39.146601500Z",
  "votes": 8,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi everyone, when I sent my first submission I was just testing some dumb idea to make matching between the datasets-labels, and the text. Surprised by getting about 82%. So I thought what if I got some other representations for each label and tried to find the matches. We only have 130 different labels. But I was shocked by the fact that there are similar datasets with the same naming. This means if we found it in the text we don't know which label should be chosen. Anyway, I built this dataset which was not helpful for name matching but I am sure that it could be helpful in the future.<br>\nGood luck with this competition and best wishes.<br>\nDataset link: <a href=\"https://www.kaggle.com/ammarali32/coleridge-external-data\" target=\"_blank\">https://www.kaggle.com/ammarali32/coleridge-external-data</a>.</p>",
  "messages": [
    {
      "id": "1250620",
      "postDate": "03/24/2021 06:54:39",
      "content": "<p>Hi everyone, when I sent my first submission I was just testing some dumb idea to make matching between the datasets-labels, and the text. Surprised by getting about 82%. So I thought what if I got some other representations for each label and tried to find the matches. We only have 130 different labels. But I was shocked by the fact that there are similar datasets with the same naming. This means if we found it in the text we don't know which label should be chosen. Anyway, I built this dataset which was not helpful for name matching but I am sure that it could be helpful in the future.<br>\nGood luck with this competition and best wishes.<br>\nDataset link: <a href=\"https://www.kaggle.com/ammarali32/coleridge-external-data\" target=\"_blank\">https://www.kaggle.com/ammarali32/coleridge-external-data</a>.</p>",
      "rawMarkdown": "Hi everyone, when I sent my first submission I was just testing some dumb idea to make matching between the datasets-labels, and the text. Surprised by getting about 82%. So I thought what if I got some other representations for each label and tried to find the matches. We only have 130 different labels. But I was shocked by the fact that there are similar datasets with the same naming. This means if we found it in the text we don't know which label should be chosen. Anyway, I built this dataset which was not helpful for name matching but I am sure that it could be helpful in the future.\nGood luck with this competition and best wishes.\nDataset link: https://www.kaggle.com/ammarali32/coleridge-external-data.",
      "votes": null
    },
    {
      "id": "1250630",
      "postDate": "03/24/2021 07:07:56",
      "content": "<ul>\n<li>List of public dataset names will definitely help in this competition (to catch unknown datasets not included in train)</li>\n<li>Since there are <a href=\"https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228322\" target=\"_blank\">very few dataset labels in train data, spelled differently</a>, matching of <a href=\"https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228247\" target=\"_blank\">key phrases</a> (known dataset labels) to text will get a good LB score</li>\n</ul>",
      "rawMarkdown": "List of public dataset names will definitely help in this competition (to catch unknown datasets not included in train)\n- Since there are [very few dataset labels in train data, spelled differently](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228322), matching of [key phrases](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228247) (known dataset labels) to text will get a good LB score",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1250630,
      "author_name": "sirishks",
      "author_url": "",
      "post_date": "03/24/2021 07:07:56",
      "content": "<ul>\n<li>List of public dataset names will definitely help in this competition (to catch unknown datasets not included in train)</li>\n<li>Since there are <a href=\"https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228322\" target=\"_blank\">very few dataset labels in train data, spelled differently</a>, matching of <a href=\"https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228247\" target=\"_blank\">key phrases</a> (known dataset labels) to text will get a good LB score</li>\n</ul>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1250620": "Hi everyone, when I sent my first submission I was just testing some dumb idea to make matching between the datasets-labels, and the text. Surprised by getting about 82%. So I thought what if I got some other representations for each label and tried to find the matches. We only have 130 different labels. But I was shocked by the fact that there are similar datasets with the same naming. This means if we found it in the text we don't know which label should be chosen. Anyway, I built this dataset which was not helpful for name matching but I am sure that it could be helpful in the future.\nGood luck with this competition and best wishes.\nDataset link: https://www.kaggle.com/ammarali32/coleridge-external-data.",
    "1250630": "List of public dataset names will definitely help in this competition (to catch unknown datasets not included in train)\n- Since there are [very few dataset labels in train data, spelled differently](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228322), matching of [key phrases](https://www.kaggle.com/c/coleridgeinitiative-show-us-the-data/discussion/228247) (known dataset labels) to text will get a good LB score"
  },
  "source": "meta"
}