README.md

elmo_experiments

This repository is experiments of ELMo that deep contextualized word representation by Keras.

ELMo is contextualized word representations using character-based word representations and bidirectional LSTMs. TensorFlow hub provides ELMo module that was trained on the 1 Billion Word Benchmark.

Getting start

$ pip install -r requirements.txt

NOTE:If use tensorflow with cpu support, need installing that by your self.

Experiments

Text classification on IMDB

This experiments is text classification with ELMo, LSTM and word embedding on IMDB dataset. Evaluate ELMo text classification and compare that to LSTM model.

https://github.com/kamujun/elmo_experiments/blob/master/elmo_experiment/notebooks/elmo_text_classification_on_imdb.ipynb

Result

ELMo's 1st epoch get best score.