b-sayah / deid_clinical_data Goto Github PK

Some Named Entity Recognition (NER) approaches applied to multiple datasets and extended for the main goal of this work which is the deidentification of clinical data of electronic health records.

Stanford NLP-NER tutorial: https://github.com/cs230-stanford/cs230-code-examples/tree/master/pytorch/nlp

Log-Linear Models, MEMMs, and CRFs: http://www.cs.columbia.edu/~mcollins/crf.pdf

BiLSTM-CRF: https://jovian.ai/abdulmajee/bilstm-crf#C29 https://pytorch.org/tutorials/beginner/nlp/advanced_tutorial.html

Transfer learning - Bert: https://colab.research.google.com/github/abhimishra91/transformers-tutorials/blob/master/transformers_ner.ipynb

Interesting tutorial, simplistic implementation: https://www.depends-on-the-definition.com/sequence-tagging-lstm-crf/

Interesting repo, advanced implementation (CONLL): https://github.com/allanj/pytorch_neural_crf/blob/master/src/model/module/linear_crf_inferencer.py

• Paper reading related to the topic (deidentification)
  • Ch-8 Sequence tagging NLP Book from Jufrasky
  • other papers and blogposts (see section Reference & interesting links to consider )
• i2b2 data request
• explore i2b2 dataset
• build code on easier/standard datasets (e.g CoNLL-2003 or Kaggle NER Dataset)
• lstm, bilstm, then bilstm-crf (consider implementation of pytorch-crf, and crf "from scratch")
• apply to i2b2
• compare to transfer learning (Bert, SciBert, Electra)
• FASTER CRF

Explore : character embedding https://gist.github.com/DuaneNielsen/4e45408948b0aca9b66b7a55ddec8950 https://pytorch.org/tutorials/intermediate/char_rnn_classification_tutorial.html

activate the environement variables : env-crf (#TODO need to generate install requirements)

python train.py --data_dir data/small --model_dir experiments/base_model

Allennlp (not sure if the code is working) https://github.com/marumalo/bilstm-crf

NER, multiple papers with code http://nlpprogress.com/english/named_entity_recognition.html

An interesting Paper with pseudocode https://arxiv.org/pdf/1508.01991.pdf reference for build upon pytorch tutorial https://github.com/jidasheng/bi-lstm-crf

more codes : https://github.com/marumalo/bilstm-crf using AllelNLP

https://github.com/jidasheng/bi-lstm-crf/tree/master/bi_lstm_crf