Named Entity Recognition (NER) using LSTM with Keras Dataset Source NER Dataset Model Used Bi-directional LSTM Bi-directional GRU Notes: Bi-directional GRU performed well. Accuracy model training loss training acc validation loss validation acc Bi-directional LSTM 0.039 0.988 0.048 0.985 Bi-directional GRU 0.034 0.989 0.045 0.986 Performance (Bi-directional GRU) Some prediction (Bi-directional GRU) get_ner('Sunday is close day for London .') Sunday B-tim London B-geo get_ner('I will go to London from Indonesia Sunday Morning .') London B-geo Indonesia B-geo Sunday B-tim Morning I-tim get_ner('U.S. President George Bush will visit the Vatican city Tomorrow Morning .') U.S. B-org President B-per George I-per Bush I-per Vatican B-org Tomorrow B-geo Morning I-org
