{
  "id": 141626,
  "title": "An experiment with LSTM networks",
  "url": "/competitions/jigsaw-multilingual-toxic-comment-classification/discussion/141626",
  "author_name": "PAB97",
  "post_date": "2020-04-06T21:39:09.100000",
  "votes": 4,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi everyone,</p>\n\n<p>Since the competition allows translated texts, I tried to use a model from the previous Jigsaw competition from Christoph Henkel. I implemented it using FastText and GloVe embeddings. </p>\n\n<p>The model can be further fine-tuned. </p>\n\n<p>I do feel a RNN model might have its place in an ensemble. At least using multilingual embeddings like Multilingual FastText.</p>\n\n<p>For my experiment with LSTM: <a href=\"https://www.kaggle.com/rftexas/gru-lstm-rnn-101\">https://www.kaggle.com/rftexas/gru-lstm-rnn-101</a></p>",
  "messages": [
    {
      "id": 799892,
      "postDate": "2020-04-06T21:39:09.100Z",
      "content": "<p>Hi everyone,</p>\n\n<p>Since the competition allows translated texts, I tried to use a model from the previous Jigsaw competition from Christoph Henkel. I implemented it using FastText and GloVe embeddings. </p>\n\n<p>The model can be further fine-tuned. </p>\n\n<p>I do feel a RNN model might have its place in an ensemble. At least using multilingual embeddings like Multilingual FastText.</p>\n\n<p>For my experiment with LSTM: <a href=\"https://www.kaggle.com/rftexas/gru-lstm-rnn-101\">https://www.kaggle.com/rftexas/gru-lstm-rnn-101</a></p>",
      "rawMarkdown": "Hi everyone,\n\nSince the competition allows translated texts, I tried to use a model from the previous Jigsaw competition from Christoph Henkel. I implemented it using FastText and GloVe embeddings. \n\nThe model can be further fine-tuned. \n \nI do feel a RNN model might have its place in an ensemble. At least using multilingual embeddings like Multilingual FastText.\n\nFor my experiment with LSTM: https://www.kaggle.com/rftexas/gru-lstm-rnn-101",
      "votes": 4
    },
    {
      "id": 801621,
      "postDate": "2020-04-08T16:27:28.517Z",
      "rawMarkdown": "",
      "votes": 1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 801621,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-04-08T16:27:28.517000",
      "content": "",
      "votes": 1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "799892": "Hi everyone,\n\nSince the competition allows translated texts, I tried to use a model from the previous Jigsaw competition from Christoph Henkel. I implemented it using FastText and GloVe embeddings. \n\nThe model can be further fine-tuned. \n \nI do feel a RNN model might have its place in an ensemble. At least using multilingual embeddings like Multilingual FastText.\n\nFor my experiment with LSTM: https://www.kaggle.com/rftexas/gru-lstm-rnn-101",
    "801621": ""
  }
}