multi-class-text-classification-cnn-rnn

Why we included this project

Most text-classification tutorials stop at a single model snippet, but this repo carries the job through to the end. It trains a hybrid TextCNN + GRU network on the Kaggle San Francisco Crime dataset, mapping free-text incident descriptions into 39 crime categories, then shows how to export the model and run predictions on new data. The train and predict scripts, saved vocabulary, and label mappings make it a workable template for anyone building a sentence classifier on their own labeled text. It is also a useful reference for teams weighing CNN against RNN (GRU/LSTM) architectures for short-text classification, since the code demonstrates how both approaches fit together in Keras.