sentence_cnn_save.py

'''
Written by Austin Walters
Last Edit: January 2, 2019
For use on austingwalters.com

A CNN to classify a sentence as one 
of the common sentance types:
Question, Statement, Command, Exclamation

Heavily Inspired by Keras Examples: 
https://github.com/keras-team/keras
'''

from __future__ import print_function

import os
import sys

import numpy as np
import keras

from sentence_types import load_encoded_data
from sentence_types import encode_data, import_embedding
from sentence_types import get_custom_test_comments

from keras.preprocessing import sequence
from keras.models import Sequential, model_from_json
from keras.layers import Dense, Dropout, Activation, Embedding
from keras.layers import Conv1D, GlobalMaxPooling1D

from keras.preprocessing.text import Tokenizer

# Use can load a different model if desired
model_name      = "models/cnn"
embedding_name  = "data/default"
load_model_flag = False
arguments       = sys.argv[1:len(sys.argv)]
if len(arguments) == 1:
    model_name = arguments[0]
    load_model_flag = os.path.isfile(model_name+".json")
print(model_name)
print("Load Model?", (load_model_flag))

# Model configuration
maxlen = 500
batch_size = 64
embedding_dims = 75
filters = 100
kernel_size = 5
hidden_dims = 350
epochs = 2

# Add parts-of-speech to data
pos_tags_flag = True


# Export & load embeddings
x_train, x_test, y_train, y_test = load_encoded_data(data_split=0.8,
                                                     embedding_name=embedding_name,
                                                     pos_tags=pos_tags_flag)

word_encoding, category_encoding = import_embedding(embedding_name)

max_words   = len(word_encoding) + 1
num_classes = np.max(y_train) + 1

print(max_words, 'words')
print(num_classes, 'classes')

print('Pad sequences (samples x time)')
x_train = sequence.pad_sequences(x_train, maxlen=maxlen)
x_test = sequence.pad_sequences(x_test, maxlen=maxlen)

print('Convert class vector to binary class matrix '
      '(for use with categorical_crossentropy)')
y_train = keras.utils.to_categorical(y_train, num_classes)
y_test = keras.utils.to_categorical(y_test, num_classes)

if not load_model_flag:

    print('Constructing model!')

    model = Sequential()

    model.add(Embedding(max_words, embedding_dims,
                        input_length=maxlen))
    
    model.add(Dropout(0.2))
    
    model.add(Conv1D(filters, kernel_size, padding='valid',
                     activation='relu', strides=1))
    
    model.add(GlobalMaxPooling1D())
    
    model.add(Dense(hidden_dims))
    model.add(Dropout(0.2))
    model.add(Activation('relu'))
    
    model.add(Dense(num_classes))
    model.add(Activation('softmax'))
    
    model.compile(loss='categorical_crossentropy',
                  optimizer='adam',
                  metrics=['accuracy'])
    
    model.fit(x_train, y_train, batch_size=batch_size,
              epochs=epochs, validation_data=(x_test, y_test))

    model_json = model.to_json()
    with open(model_name + ".json", "w") as json_file:
        json_file.write(model_json)
    
    # serialize weights to HDF5
    model.save_weights(model_name + ".h5")
    print("Saved model to disk")

else:

    print('Loading model!')

    # load json and create model
    json_file = open(model_name + '.json', 'r')
    loaded_model_json = json_file.read()
    json_file.close()
    model = model_from_json(loaded_model_json)
    
    # load weights into new model
    model.load_weights(model_name + ".h5")
    print("Loaded model from disk")
    
    # evaluate loaded model on test data
    model.compile(loss='categorical_crossentropy',
                  optimizer='adam',
                  metrics=['accuracy'])


score = model.evaluate(x_test, y_test, batch_size=batch_size, verbose=1)

print('Test accuracy:', score[1])

test_comments, test_comments_category = get_custom_test_comments()

x_test, _, y_test, _ = encode_data(test_comments, test_comments_category,
                                   data_split=1.0,
                                   embedding_name=embedding_name,
                                   add_pos_tags_flag=pos_tags_flag)

x_test = sequence.pad_sequences(x_test, maxlen=maxlen)
y_test = keras.utils.to_categorical(y_test, num_classes)

score = model.evaluate(x_test, y_test,
                       batch_size=batch_size, verbose=1)


print('Manual test')
print('Test accuracy:', score[1])

# Show predictions
print(len(x_test))
predictions = model.predict(x_test, batch_size=batch_size, verbose=1)

real = []
test = []
for i in range(0, len(predictions)):
    real.append(y_test[i].argmax(axis=0))
    test.append(predictions[i].argmax(axis=0))

print("Predictions")
print("Real", real)
print("Test", test)