This is an implementation of a Convolutional Neural Network for Text Classification in Tensorflow. The model used above is a special type of CNN in which convolution is performed on the input three times with different filters and then combined together and is followed by a fully connected output layer.
The model performed well when tested on IMDB's movie reviews dataset. The model was trained to classify weather a review is positive or negative
To train the model run the following command
python classifier.py --image_path <path-to-image-file>When training a model will be saved inside the runs folder at each 100th step. To use the trained model call the function classify inside textclassifier file
import text_classifier
result = text_classifier.classify(checkpoint_dir, x_text)where checkpoint_dir is the directory in which the model are saved and x_text is the text to be classified. (refer usage_example.py for an example of the usage)