NER
A lightweight, trainable, Keras-based Named Entity Recognizer
A model that accepts a named entity recognition (NER) training file in .iob format and trains an LSTM model to predict the IOB tags for each token of each sentence. All sentences are pre-padded to be of the same length. Uses an internal validation set to train with early stopping and also evaluates model performance on a test set. The weights and configuration of the trained model are also saved out.
This repo currently trains on the MIT trivia10k13 Movie Corpus.