import numpy as np
import pandas as pd
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from nltk.stem.lancaster import LancasterStemmer
import nltk
import re
from sklearn.preprocessing import OneHotEncoder
import matplotlib.pyplot as plt
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from keras.utils import to_categorical
from keras.models import Sequential, load_model
from keras.layers import Dense, GRU, LSTM, Bidirectional, Embedding, Dropout
from keras.callbacks import ModelCheckpoint
from livelossplot.tf_keras import PlotLossesCallback
from livelossplot import PlotLossesKerasDataset link:
def load_dataset(filename):
df = pd.read_csv(filename)
label = df["label"]
unique_label = list(set(label))
sentences = list(df["text"])
return (df, label, unique_label, sentences)df, label, unique_label, sentences = load_dataset('iseardataset.csv')print(unique_label)['fear', 'anger', 'shame', 'sadness', 'joy', 'disgust', 'guilt']
print(df.head(10)) label text Unnamed: 2
0 joy On days when I feel close to my partner and ot... NaN
1 fear Every time I imagine that someone I love or I ... NaN
2 anger When I had been obviously unjustly treated and... NaN
3 sadness When I think about the short time that we live... NaN
4 disgust At a gathering I found myself involuntarily si... NaN
5 shame When I realized that I was directing the feeli... NaN
6 guilt I feel guilty when when I realize that I consi... NaN
7 joy After my girlfriend had taken her exam we went... NaN
8 fear When, for the first time I realized the meanin... NaN
9 anger When a car is overtaking another and I am forc... NaN
import seaborn as sns
import tkinter
import matplotlib
matplotlib.use('agg')
import matplotlib.pyplot as plt
%matplotlib inline
sns.countplot(x="label", data=df)<AxesSubplot:xlabel='label', ylabel='count'>
print(sentences[:5])['On days when I feel close to my partner and other friends. \nWhen I feel at peace with myself and also experience a close \ncontact with people whom I regard greatly.', 'Every time I imagine that someone I love or I could contact a \nserious illness, even death.', 'When I had been obviously unjustly treated and had no possibility \nof elucidating this.', 'When I think about the short time that we live and relate it to \nthe periods of my life when I think that I did not use this \nshort time.', 'At a gathering I found myself involuntarily sitting next to two \npeople who expressed opinions that I considered very low and \ndiscriminating.']
nltk.download("stopwords")
nltk.download("punkt")[nltk_data] Downloading package stopwords to
[nltk_data] /home/shiningflash/nltk_data...
[nltk_data] Package stopwords is already up-to-date!
[nltk_data] Downloading package punkt to
[nltk_data] /home/shiningflash/nltk_data...
[nltk_data] Package punkt is already up-to-date!
True
#define stemmer
stemmer = LancasterStemmer()def cleaning(sentences):
words = []
for s in sentences:
clean = re.sub(r'[^ a-z A-Z 0-9]', " ", s)
w = word_tokenize(clean)
words.append([i.lower() for i in w])
return words cleaned_words = cleaning(sentences)
print(len(cleaned_words))
print(cleaned_words[:2]) 7516
[['on', 'days', 'when', 'i', 'feel', 'close', 'to', 'my', 'partner', 'and', 'other', 'friends', 'when', 'i', 'feel', 'at', 'peace', 'with', 'myself', 'and', 'also', 'experience', 'a', 'close', 'contact', 'with', 'people', 'whom', 'i', 'regard', 'greatly'], ['every', 'time', 'i', 'imagine', 'that', 'someone', 'i', 'love', 'or', 'i', 'could', 'contact', 'a', 'serious', 'illness', 'even', 'death']]
def create_tokenizer(words, filters = '!"#$%&()*+,-./:;<=>?@[\]^_`{|}~'):
token = Tokenizer(filters = filters)
token.fit_on_texts(words)
return tokendef max_length(words):
return(len(max(words, key = len)))word_tokenizer = create_tokenizer(cleaned_words)
vocab_size = len(word_tokenizer.word_index) + 1
max_length = max_length(cleaned_words)
print("Vocab Size = %d and Maximum length = %d" % (vocab_size, max_length))Vocab Size = 8989 and Maximum length = 179
def encoding_doc(token, words):
return(token.texts_to_sequences(words))encoded_doc = encoding_doc(word_tokenizer, cleaned_words)def padding_doc(encoded_doc, max_length):
return(pad_sequences(encoded_doc, maxlen = max_length, padding = "post"))padded_doc = padding_doc(encoded_doc, max_length)print("Shape of padded docs = ",padded_doc.shape)Shape of padded docs = (7516, 179)
#tokenizer with filter changed
output_tokenizer = create_tokenizer(unique_label, filters = '!"#$%&()*+,-/:;<=>?@[\]^`{|}~')output_tokenizer.word_index{'fear': 1,
'anger': 2,
'shame': 3,
'sadness': 4,
'joy': 5,
'disgust': 6,
'guilt': 7}
encoded_output = encoding_doc(output_tokenizer, label)encoded_output = np.array(encoded_output).reshape(len(encoded_output), 1)encoded_output.shape(7516, 1)
def one_hot(encode):
o = OneHotEncoder(sparse = False)
return(o.fit_transform(encode))output_one_hot = one_hot(encoded_output)output_one_hot.shape(7516, 7)
from sklearn.model_selection import train_test_splittrain_X, val_X, train_Y, val_Y = train_test_split(padded_doc, output_one_hot, shuffle = True, test_size = 0.2)print("Shape of train_X = %s and train_Y = %s" % (train_X.shape, train_Y.shape))
print("Shape of val_X = %s and val_Y = %s" % (val_X.shape, val_Y.shape))Shape of train_X = (6012, 179) and train_Y = (6012, 7)
Shape of val_X = (1504, 179) and val_Y = (1504, 7)
def create_model(vocab_size, max_length):
model = Sequential()
model.add(Embedding(vocab_size, 128, input_length = max_length, trainable = False))
model.add(Bidirectional(GRU(128)))
model.add(Dense(32, activation = "relu"))
model.add(Dropout(0.5))
model.add(Dense(7, activation = "softmax"))
return modelmodel = create_model(vocab_size, max_length)
model.compile(loss = "categorical_crossentropy", optimizer = "adam", metrics = ["accuracy"])
model.summary()Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 179, 128) 1150592
_________________________________________________________________
bidirectional (Bidirectional (None, 256) 198144
_________________________________________________________________
dense (Dense) (None, 32) 8224
_________________________________________________________________
dropout (Dropout) (None, 32) 0
_________________________________________________________________
dense_1 (Dense) (None, 7) 231
=================================================================
Total params: 1,357,191
Trainable params: 206,599
Non-trainable params: 1,150,592
_________________________________________________________________
filename = 'model.h5'
checkpoint = ModelCheckpoint(filename,
monitor='val_loss',
verbose=1,
save_best_only=True,
mode='min')hist = model.fit(train_X, train_Y,
epochs = 100,
batch_size = 32,
validation_data = (val_X, val_Y),
callbacks = [PlotLossesKeras(), checkpoint])
accuracy
training (min: 0.143, max: 0.936, cur: 0.936)
validation (min: 0.186, max: 0.449, cur: 0.427)
Loss
training (min: 0.173, max: 1.947, cur: 0.173)
validation (min: 1.556, max: 6.043, cur: 6.043)
Epoch 00100: val_loss did not improve from 1.55581
188/188 [==============================] - 28s 147ms/step - loss: 0.1726 - accuracy: 0.9356 - val_loss: 6.0430 - val_accuracy: 0.4269
def create_model(vocab_size, max_length):
model = Sequential()
model.add(Embedding(vocab_size, 128, input_length = max_length, trainable = False))
model.add(Bidirectional(LSTM(128)))
model.add(Dense(32, activation = "relu"))
model.add(Dropout(0.5))
model.add(Dense(7, activation = "softmax"))
return model
model_lstm = create_model(vocab_size, max_length)
model_lstm.compile(loss = "categorical_crossentropy", optimizer = "adam", metrics = ["accuracy"])
model_lstm.summary()Model: "sequential_1"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding_1 (Embedding) (None, 179, 128) 1150592
_________________________________________________________________
bidirectional_1 (Bidirection (None, 256) 263168
_________________________________________________________________
dense_2 (Dense) (None, 32) 8224
_________________________________________________________________
dropout_1 (Dropout) (None, 32) 0
_________________________________________________________________
dense_3 (Dense) (None, 7) 231
=================================================================
Total params: 1,422,215
Trainable params: 271,623
Non-trainable params: 1,150,592
_________________________________________________________________
filename = 'model_lstm.h5'
checkpoint = ModelCheckpoint(filename,
monitor='val_loss',
verbose=1,
save_best_only=True,
mode='min')
hist = model_lstm.fit(train_X, train_Y,
epochs = 100,
batch_size = 32,
validation_data = (val_X, val_Y),
callbacks = [PlotLossesKeras(), checkpoint])
accuracy
training (min: 0.143, max: 0.875, cur: 0.867)
validation (min: 0.158, max: 0.437, cur: 0.430)
Loss
training (min: 0.350, max: 1.947, cur: 0.379)
validation (min: 1.602, max: 3.759, cur: 3.562)
Epoch 00100: val_loss did not improve from 1.60191
188/188 [==============================] - 46s 247ms/step - loss: 0.3785 - accuracy: 0.8674 - val_loss: 3.5621 - val_accuracy: 0.4295
model = load_model("model_lstm.h5")def predictions(text):
clean = re.sub(r'[^ a-z A-Z 0-9]', " ", text)
test_word = word_tokenize(clean)
test_word = [w.lower() for w in test_word]
test_ls = word_tokenizer.texts_to_sequences(test_word)
if [] in test_ls:
test_ls = list(filter(None, test_ls))
test_ls = np.array(test_ls).reshape(1, len(test_ls))
x = padding_doc(test_ls, max_length)
pred = model.predict(x)
return preddef get_final_output(pred, classes):
predictions = pred[0]
classes = np.array(classes)
ids = np.argsort(-predictions)
classes = classes[ids]
predictions = -np.sort(-predictions)
for i in range(pred.shape[1]):
print("%s has confidence = %s" % (classes[i], (predictions[i])))
return classes[0]def get_emotion(text):
pred = predictions(text)
result = get_final_output(pred, unique_label)
print('\nans: {}\n'.format(result))get_emotion("I did not help out enough at my thesis team.")guilt has confidence = 0.49738222
shame has confidence = 0.36354083
anger has confidence = 0.05595107
sadness has confidence = 0.037771154
fear has confidence = 0.022115987
disgust has confidence = 0.015596252
joy has confidence = 0.0076424023
ans: guilt
get_emotion("When someone stole my bike.")anger has confidence = 0.21699908
fear has confidence = 0.16840756
guilt has confidence = 0.16005495
sadness has confidence = 0.15744306
shame has confidence = 0.14280295
disgust has confidence = 0.11216525
joy has confidence = 0.042127196
ans: anger
get_emotion("When my girlfriend left me and tell me that I am not fit for her.")anger has confidence = 0.40398777
sadness has confidence = 0.19539253
disgust has confidence = 0.126978
guilt has confidence = 0.090000965
joy has confidence = 0.08952445
shame has confidence = 0.070161685
fear has confidence = 0.023954567
ans: anger
get_emotion("During the Christmas holidays, I met some of my old friends.")joy has confidence = 0.3547267
disgust has confidence = 0.2273542
anger has confidence = 0.15483476
shame has confidence = 0.105804436
sadness has confidence = 0.066970326
guilt has confidence = 0.060981136
fear has confidence = 0.029328424
ans: joy