ruankie/spam-detector - GitHub

Natural Language Processing with LSTM in TensorFlow 2

Project Overview

This model uses a binary classifier to classify a given message as either spam or ham (not spam). It was created using TensorFlow 2.x and consists of a Long Short Term Memory (LSTM) neural network that is able to capture long-term dependencies in the words that make up a message. This model also uses an embedding layer that encodes words into an abstract 20-dimensional space so that they can be represented as vectors with contextual proximity.

After training this model for 10 epochs on a training data set of 3,715 messages (70% of data set), it produced a validation-accuracy of around 87%.

References

This project was based on a project in this TensorFlow course on Udemy by the Lazy Programmer. The data set used can be found here.