klrshak / sms-spam-detector-nlp

Natural Language Processing using Tensorflow, the model is trained on >5000 SMS text messages to identify spam messages with an validation accuracy of over 98%.

Spam SMS Dectector

An example of Natural Language Processing using Tensorflow, the model is trained on >5000 SMS text messages to identify spam messages with an validation accuracy of over 98%. The dataset used is from kaggle: https://www.kaggle.com/uciml/sms-spam-collection-dataset .

Information about Dataset: The SMS Spam Collection is a set of SMS tagged messages that have been collected for SMS Spam research. It contains one set of SMS messages in English of 5,574 messages, tagged acording being ham (legitimate) or spam.

Visualising the Data

Meta and Vecs files can be useed to visualize the embeddings using TensorFlow Embedding Projector

Using Embedding Projector, As you can see in the images below certain words are given more weight as spam by the model during its learning Phase, which helps the model correctly categorize the messages

Demo Image :-