Hello!
\n\nI was hoping to get some recommendations on how you learned to do deep learning NLP and/or how you were able to get a job doing deep learning NLP.
\n\nMy background is that I'm a working data scientist with an MS in data science. My degree did not cover deep learning very much, but I've taken Ng's Deep Learning Specialization on Coursera and a few Datacamp classes on Keras & RNNs. However, I feel like I don't quite have the knowledge or experience to tackle this Kaggle competition.
\n\nWhat would you recommend? Thanks!
", "messages": [ { "id": 840041, "postDate": "2020-05-09T18:21:49.063Z", "content": "Hello Kyle, I felt like you a few months ago while starting this competition. I guess the density of DL and NLP in particular can be quite scary in the beginning. I did some cheat sheets and followed Stanford CS course. I also did a structured mind map for NLP, which is a great way I guess to have a broad overview of the field.
", "rawMarkdown": "Hello Kyle, I felt like you a few months ago while starting this competition. I guess the density of DL and NLP in particular can be quite scary in the beginning. I did some cheat sheets and followed Stanford CS course. I also did a structured mind map for NLP, which is a great way I guess to have a broad overview of the field. ", "votes": 2 }, { "id": 835030, "postDate": "2020-05-06T02:04:08.987Z", "content": "Hello!
\n\nI was hoping to get some recommendations on how you learned to do deep learning NLP and/or how you were able to get a job doing deep learning NLP.
\n\nMy background is that I'm a working data scientist with an MS in data science. My degree did not cover deep learning very much, but I've taken Ng's Deep Learning Specialization on Coursera and a few Datacamp classes on Keras & RNNs. However, I feel like I don't quite have the knowledge or experience to tackle this Kaggle competition.
\n\nWhat would you recommend? Thanks!
", "rawMarkdown": "Hello!\n\nI was hoping to get some recommendations on how you learned to do deep learning NLP and/or how you were able to get a job doing deep learning NLP.\n\nMy background is that I'm a working data scientist with an MS in data science. My degree did not cover deep learning very much, but I've taken Ng's Deep Learning Specialization on Coursera and a few Datacamp classes on Keras & RNNs. However, I feel like I don't quite have the knowledge or experience to tackle this Kaggle competition.\n\nWhat would you recommend? Thanks!", "votes": 2 }, { "id": 839390, "postDate": "2020-05-09T11:12:49.153Z", "content": "I think you have enough knowledge required to start of with this competition.\nAndrew Ng's course is good enough to get a deep understanding of the various sequence models (RNNs, LSTMs, GRU, Attention with Sequence Models). \nThe course do not cover Transformers, which you can learn from this excellent blog.\nOnce you know about transformers, next thing to learn about is BERT.
\n\nIf you are finding it difficult to setup your own code from stratch, then you can refer to various public notebooks which are published.\nYou would find notebooks which use either Tensorflow or Pytorch to develop the model. Pytorch would be a bit difficult to grasp as compared to tensorflow. Refer @xhlulu kernel- Jigsaw TPU: XLM-Roberta. The kernel is quite clean and easy to understand.
\n\nHope this helps :)
", "rawMarkdown": "I think you have enough knowledge required to start of with this competition.\nAndrew Ng's course is good enough to get a deep understanding of the various sequence models (RNNs, LSTMs, GRU, Attention with Sequence Models). \nThe course do not cover Transformers, which you can learn from this [excellent blog](http://jalammar.github.io/illustrated-transformer/).\nOnce you know about transformers, next thing to learn about is [BERT](http://jalammar.github.io/illustrated-bert/). \n\nIf you are finding it difficult to setup your own code from stratch, then you can refer to various public notebooks which are published.\nYou would find notebooks which use either Tensorflow or Pytorch to develop the model. Pytorch would be a bit difficult to grasp as compared to tensorflow. Refer @xhlulu kernel- [Jigsaw TPU: XLM-Roberta](https://www.kaggle.com/xhlulu/jigsaw-tpu-xlm-roberta). The kernel is quite clean and easy to understand. \n\nHope this helps :)\n" }, { "id": 837013, "postDate": "2020-05-07T12:49:24.340Z", "content": "The most basic NLP is classifying text. This competition requires selecting portions from text which is a little more involved but not that much more difficult to learn. There are many wonderful starter notebooks to get you started
", "rawMarkdown": "The most basic NLP is classifying text. This competition requires selecting portions from text which is a little more involved but not that much more difficult to learn. There are many wonderful starter notebooks to get you started", "votes": -2 }, { "id": 839052, "postDate": "2020-05-09T03:09:22.833Z", "rawMarkdown": "", "votes": -2, "isDeleted": true }, { "id": 839051, "postDate": "2020-05-09T03:06:53.697Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 840041, "author_name": "PAB97", "author_url": "", "post_date": "2020-05-09T18:21:49.063000", "content": "Hello Kyle, I felt like you a few months ago while starting this competition. I guess the density of DL and NLP in particular can be quite scary in the beginning. I did some cheat sheets and followed Stanford CS course. I also did a structured mind map for NLP, which is a great way I guess to have a broad overview of the field.
", "votes": 2, "replies": [] }, { "id": 839390, "author_name": "sajwankit", "author_url": "", "post_date": "2020-05-09T11:12:49.153000", "content": "I think you have enough knowledge required to start of with this competition.\nAndrew Ng's course is good enough to get a deep understanding of the various sequence models (RNNs, LSTMs, GRU, Attention with Sequence Models). \nThe course do not cover Transformers, which you can learn from this excellent blog.\nOnce you know about transformers, next thing to learn about is BERT.
\n\nIf you are finding it difficult to setup your own code from stratch, then you can refer to various public notebooks which are published.\nYou would find notebooks which use either Tensorflow or Pytorch to develop the model. Pytorch would be a bit difficult to grasp as compared to tensorflow. Refer @xhlulu kernel- Jigsaw TPU: XLM-Roberta. The kernel is quite clean and easy to understand.
\n\nHope this helps :)
", "votes": 0, "replies": [] }, { "id": 837013, "author_name": "Ankit Maurya", "author_url": "", "post_date": "2020-05-07T12:49:24.340000", "content": "The most basic NLP is classifying text. This competition requires selecting portions from text which is a little more involved but not that much more difficult to learn. There are many wonderful starter notebooks to get you started
", "votes": -2, "replies": [] }, { "id": 839052, "author_name": "", "author_url": "", "post_date": "2020-05-09T03:09:22.833000", "content": "", "votes": -2, "replies": [] }, { "id": 839051, "author_name": "", "author_url": "", "post_date": "2020-05-09T03:06:53.697000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "840041": "Hello Kyle, I felt like you a few months ago while starting this competition. I guess the density of DL and NLP in particular can be quite scary in the beginning. I did some cheat sheets and followed Stanford CS course. I also did a structured mind map for NLP, which is a great way I guess to have a broad overview of the field. ", "835030": "Hello!\n\nI was hoping to get some recommendations on how you learned to do deep learning NLP and/or how you were able to get a job doing deep learning NLP.\n\nMy background is that I'm a working data scientist with an MS in data science. My degree did not cover deep learning very much, but I've taken Ng's Deep Learning Specialization on Coursera and a few Datacamp classes on Keras & RNNs. However, I feel like I don't quite have the knowledge or experience to tackle this Kaggle competition.\n\nWhat would you recommend? Thanks!", "839390": "I think you have enough knowledge required to start of with this competition.\nAndrew Ng's course is good enough to get a deep understanding of the various sequence models (RNNs, LSTMs, GRU, Attention with Sequence Models). \nThe course do not cover Transformers, which you can learn from this [excellent blog](http://jalammar.github.io/illustrated-transformer/).\nOnce you know about transformers, next thing to learn about is [BERT](http://jalammar.github.io/illustrated-bert/). \n\nIf you are finding it difficult to setup your own code from stratch, then you can refer to various public notebooks which are published.\nYou would find notebooks which use either Tensorflow or Pytorch to develop the model. Pytorch would be a bit difficult to grasp as compared to tensorflow. Refer @xhlulu kernel- [Jigsaw TPU: XLM-Roberta](https://www.kaggle.com/xhlulu/jigsaw-tpu-xlm-roberta). The kernel is quite clean and easy to understand. \n\nHope this helps :)\n", "837013": "The most basic NLP is classifying text. This competition requires selecting portions from text which is a little more involved but not that much more difficult to learn. There are many wonderful starter notebooks to get you started", "839052": "", "839051": "" } }