{
  "id": 229195,
  "title": "Notes about points of view on the task of this competition",
  "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/229195",
  "author_name": "Anaida Shaginian",
  "post_date": "2021-03-28T20:40:29.971000",
  "votes": 5,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Here I can share some my thoughts about how we can look at the problem of this competition.</p>\n<p>1) Classification task. We can solve here the classification problem and classify the texts to the datasets which are mentioned in texts. It's easy but the minus is that we are bounded by the training dataset and actually don't know what we can see in the test set. Also we wouldn/t be able to classify one object to several classes.</p>\n<p>2) Named Entity Recognition. The minus is that we see the exact matching in a small number of objects (<a href=\"https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt)\" target=\"_blank\">https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt)</a>. But we cat use this if we improve the approach (for example, wil try not to find the exact matching but similar text using different similarity methods - cosine, jaccard, etc)</p>\n<p>3) RNN (LSTM). Generate next sentence (the dataset name).</p>\n<p>4) Question-Answering.</p>\n<p>Any other ideas for this competition?</p>",
  "messages": [
    {
      "id": 1255469,
      "postDate": "2021-03-28T20:40:29.970Z",
      "content": "<p>Here I can share some my thoughts about how we can look at the problem of this competition.</p>\n<p>1) Classification task. We can solve here the classification problem and classify the texts to the datasets which are mentioned in texts. It's easy but the minus is that we are bounded by the training dataset and actually don't know what we can see in the test set. Also we wouldn/t be able to classify one object to several classes.</p>\n<p>2) Named Entity Recognition. The minus is that we see the exact matching in a small number of objects (<a href=\"https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt)\" target=\"_blank\">https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt)</a>. But we cat use this if we improve the approach (for example, wil try not to find the exact matching but similar text using different similarity methods - cosine, jaccard, etc)</p>\n<p>3) RNN (LSTM). Generate next sentence (the dataset name).</p>\n<p>4) Question-Answering.</p>\n<p>Any other ideas for this competition?</p>",
      "rawMarkdown": "Here I can share some my thoughts about how we can look at the problem of this competition.\n\n1) Classification task. We can solve here the classification problem and classify the texts to the datasets which are mentioned in texts. It's easy but the minus is that we are bounded by the training dataset and actually don't know what we can see in the test set. Also we wouldn/t be able to classify one object to several classes.\n\n2) Named Entity Recognition. The minus is that we see the exact matching in a small number of objects (https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt). But we cat use this if we improve the approach (for example, wil try not to find the exact matching but similar text using different similarity methods - cosine, jaccard, etc)\n\n3) RNN (LSTM). Generate next sentence (the dataset name).\n\n4) Question-Answering.\n\nAny other ideas for this competition?",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1255469": "Here I can share some my thoughts about how we can look at the problem of this competition.\n\n1) Classification task. We can solve here the classification problem and classify the texts to the datasets which are mentioned in texts. It's easy but the minus is that we are bounded by the training dataset and actually don't know what we can see in the test set. Also we wouldn/t be able to classify one object to several classes.\n\n2) Named Entity Recognition. The minus is that we see the exact matching in a small number of objects (https://www.kaggle.com/anaidashaginian/named-entity-recognition-with-spacy-first-attempt). But we cat use this if we improve the approach (for example, wil try not to find the exact matching but similar text using different similarity methods - cosine, jaccard, etc)\n\n3) RNN (LSTM). Generate next sentence (the dataset name).\n\n4) Question-Answering.\n\nAny other ideas for this competition?"
  }
}