/blstm_adr

For this problem, we proposed the use of bidirectional-LSTM’s(Long Short Term Memory) with 1-D CNN layer to classify patient notes at character level and at word level. The 1-D CNN is employed to scale back the training time. In order to improve the performance, we will also feed the network combined word embedding consisting of Pre-trained word2vec 100 dimension word embedding trained on the Twitter ADR Dataset database and character embedding generated by a Char-CNN for Named Entity Recognition

Primary LanguageJupyter Notebook

Stargazers

No one’s star this repository yet.