LSTM Spam Detector

Spam Prediction on SMS Dataset

Models Tested

• Tensorflow LSTM

Word Embeddings Tested

Test Data was composed of 25% of entire dataset.

GloVe

• Test Accuracy: 95.6%

BERT

• Test Accuracy: 99.09%

Data

https://archive.ics.uci.edu/ml/datasets/SMS+Spam+Collection

Word Embeddings

https://nlp.stanford.edu/projects/glove/

https://github.com/google-research/bert