https://github.com/mxhofer/Named-Entity-Recognition-BidirectionalLSTM-CNN-CoNLL

Keras implementation of "Few-shot Learning for Named Entity Recognition in Medical Text"
https://github.com/mxhofer/Named-Entity-Recognition-BidirectionalLSTM-CNN-CoNLL

bidirectional-lstm cnn-keras conll-2003 keras

Last synced: 2 months ago
JSON representation

Keras implementation of "Few-shot Learning for Named Entity Recognition in Medical Text"

• Host: GitHub
• URL: https://github.com/mxhofer/Named-Entity-Recognition-BidirectionalLSTM-CNN-CoNLL
• Owner: mxhofer
• License: mit
• Created: 2018-06-02T11:04:52.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2019-09-15T12:27:20.000Z (over 5 years ago)
• Last Synced: 2024-08-03T17:14:29.466Z (6 months ago)
• Topics: bidirectional-lstm, cnn-keras, conll-2003, keras
• Language: Jupyter Notebook
• Homepage: https://arxiv.org/abs/1811.05468
• Size: 9.91 MB
• Stars: 177
• Watchers: 11
• Forks: 83
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

        
# Named-Entity-Recognition-BLSTM-CNN-CoNLL

  Keras implementation of the Bidirectional LSTM and CNN model similar to Chiu and Nichols (2016) for CoNLL 2003 news data. Paper: https://arxiv.org/abs/1811.05468

The implementation differs from the original paper in these ways:

  1) no lexicons

  2) Nadam optimizer used instead of SGD

  3) Parameters: LSTM cell size of 200 (vs 275), dropout of 0.5 (vs 0.68)

Here is the corresponding Medium post with more details: https://medium.com/@maxhofer/deep-learning-for-named-entity-recognition-2-implementing-the-state-of-the-art-bidirectional-lstm-4603491087f1

Code adapted from: https://github.com/kamalkraj/Named-Entity-Recognition-with-Bidirectional-LSTM-CNNs

# Result 

  The implementation achieves a test F1 score of ~86 with 30 epochs. Increase the number of epochs to 80 reach an F1 over 90. The score produced in Chiu and Nichols (2016) is 91.62. 

# Dataset

  CoNLL-2003 newswire articles: https://www.clips.uantwerpen.be/conll2003/ner/

  GloVe vector representation from Jeffrey Pennington, Richard Socher, and Christopher D. Manning. 2014. See https://nlp.stanford.edu/projects/glove/

# Dependencies 

    1) numpy 1.15.4

    2) Keras 2.1.6

    3) Tensorflow 1.8.0

    4) Stanford GloVE embeddings