## All purpose
import pandas as pd
import numpy as np
## Visualization
import matplotlib.pyplot as plt
import seaborn as sns
import plotly
import plotly.express as px
import plotly.graph_objects as go
from wordcloud import WordCloud
## NLP
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize
from nltk.stem import WordNetLemmatizer
from nltk.tokenize import TweetTokenizer
from gensim.summarization import keywords
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
## Deeplearning/ML
import tensorflow as tf
from tensorflow.keras.preprocessing.text import Tokenizer
from sklearn.model_selection import train_test_split
import tensorflow.keras.layers as L
from tensorflow.keras.losses import MeanAbsoluteError
from tensorflow.keras.losses import SparseCategoricalCrossentropy
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.utils import plot_model
from tensorflow.keras.preprocessing.sequence import pad_sequences
## Creating sentimental polarity
analyzer = SentimentIntensityAnalyzer()
def compound_score(txt):
return analyzer.polarity_scores(txt)["compound"]
## Sentiments
def sentiment(score):
emotion = ""
if score >= 0.5:
emotion = "Positive"
elif score <= -0.5:
emotion = "Negative"
else:
emotion = "Neutral"
return emotion
Applying functions
## Importing CSV file
df = pd.read_csv("tripadvisor_hotel_reviews.csv")
## Applying Compund score
polarity_scores = df["Review"].astype("str").apply(compound_score)
df["Sentiment_Score"] = polarity_scores
## Applying Sentiment
df["Sentiment"] = df["Sentiment_Score"].apply(sentiment)
Saving file
## Saving preprocessed file
df.to_csv("Trip-Advisor-rating-sentiments.csv",index=False)
df.Sentiment.value_counts()
Loading Preprocessed Dataset
# Importing the Trip-Advisor-Hotel-Review Dataset
data=pd.read_csv('Trip-Advisor-rating-sentiments.csv')
# Having a look at the data
data.head()
data.isna().sum()
sns.countplot(data=data,x="Sentiment",palette="pastel");
Visualization
# Preparing data for visualization
Viz_1 = data[['Rating','Sentiment']].value_counts().rename_axis(['Rating','Sentiment']).reset_index(name='counts')
# Plotting the Bar Graph
fig = px.bar(x=Viz_1.Rating, y=Viz_1.counts, color=Viz_1.Sentiment,color_discrete_sequence=px.colors.qualitative.Pastel,title="Sentiment & Ratings",labels={'x':'Ratings','y':'Total Number'})
fig.show()
#Viz2 Data preparation
Viz_2 = data['Rating'].value_counts().rename_axis(['Rating']).reset_index(name='counts')
# Plotting pie chart for ratings
fig_pie = px.pie(values=Viz_2.counts, names=Viz_2.Rating, title='Rating Distribution of the data',color_discrete_sequence=px.colors.qualitative.Pastel)
fig_pie.show()
# Jointplot on the basis of Rating and Sentiment Score of the data
jp = sns.jointplot(data=data,x='Rating',y='Sentiment_Score',kind="reg",color='#ff7373')
# jp.fig.suptitle('Jointplot on the basis of Rating and Sentiment Score of the data',fontsize=20);
fig = go.Figure()
Ratings = [1,2,3,4,5]
for rating in Ratings:
fig.add_trace(go.Violin(x=data['Rating'][data['Rating'] == rating],
y=data['Sentiment_Score'][data['Rating'] == rating],
name=rating,
box_visible=True,
meanline_visible=True))
fig.update_layout(
title="Violin plot of Rating and Sentiment Score with box plot",
xaxis_title="Rating",
yaxis_title="Sentiment Score",
font=dict(
family="Courier New, monospace",
size=12,
)
)
fig.show()
Sentiment
text1 =''
for i in data[data['Sentiment']==str(Sentiment)]['Review'].values:
text1+=i + ' '
wc = WordCloud(width = 800, height = 800,background_color="white",min_font_size = 10,\
repeat=True,)
wc.generate(text1)
plt.figure(figsize = (8, 8), facecolor = None)
plt.axis("off")
plt.imshow(wc, interpolation="bilinear")
plt.title(Sentiment+' Reviews',fontsize=32);
# Getting all the reviews termed positive in a single string and forming a word cloud of the string
fig, (ax1, ax2, ax3) = plt.subplots(1, 3, figsize=[14, 14], facecolor = None)
text1 =''
for i in data[data['Sentiment']=='Positive']['Review'].values:
text1+=i + ' '
x, y = np.ogrid[:300, :300]
mask = (x - 150) ** 2 + (y - 150) ** 2 > 130 ** 2
mask = 255 * mask.astype(int)
wc1 = WordCloud(width = 800, height = 800,background_color="white",min_font_size = 10,\
repeat=True, mask=mask)
wc1.generate(text1)
ax1.axis("off")
ax1.imshow(wc1, interpolation="bilinear")
ax1.set_title('Positive Reviews',fontsize=20);
text2 =''
for i in data[data['Sentiment']=='Negative']['Review'].values:
text2+=i + ' '
x, y = np.ogrid[:300, :300]
mask = (x - 150) ** 2 + (y - 150) ** 2 > 130 ** 2
mask = 255 * mask.astype(int)
wc2 = WordCloud(width = 800, height = 800,background_color="white",min_font_size = 10,\
repeat=True, mask=mask)
wc2.generate(text2)
ax2.axis("off")
ax2.imshow(wc2, interpolation="bilinear")
ax2.set_title('Neutral Reviews',fontsize=20);
text3 =''
for i in data[data['Sentiment']=='Neutral']['Review'].values:
text3+=i + ' '
x, y = np.ogrid[:300, :300]
mask = (x - 150) ** 2 + (y - 150) ** 2 > 130 ** 2
mask = 255 * mask.astype(int)
wc3 = WordCloud(width = 800, height = 800,background_color="white",min_font_size = 10,\
repeat=True, mask=mask)
wc3.generate(text3)
ax3.axis("off")
ax3.imshow(wc3, interpolation="bilinear")
ax3.set_title('Negetive Reviews',fontsize=20);
plt.show()
input_1
keywords(input_1).split("\n")
data["keywords"] = data["Review"].apply(keywords)
data["keywords"] = data["keywords"].astype("str").str.replace('\n',',',)
words = []
for x in data.keywords.values:
x=x.split(",")
for i in x:
words.append(i)
from collections import Counter
word_counter = Counter(words)
word_df = pd.DataFrame(np.array(list(word_counter.items())),columns=["keyword","count"])
word_df["count"] = word_df["count"].astype(int)
word_df = word_df.sort_values(['count'], ascending=False)
top_20 = word_df[0:19]
word_df.head(10)
sns.set(rc={'figure.figsize':(15,8)})
fig, ax = plt.subplots()
ax = sns.barplot(data=top_20,x="keyword",y="count",palette="pastel")
ax.patch.set_visible(False)
ax.tick_params(axis='x', labelrotation = 45)
ax.set_title("Top 20 Keywords",fontsize=20);
import nltk
nltk.download('stopwords')
nltk.download('wordnet')
nltk.download('punkt')
[nltk_data] Downloading package stopwords to /root/nltk_data...
[nltk_data] Package stopwords is already up-to-date!
[nltk_data] Downloading package wordnet to /root/nltk_data...
[nltk_data] Package wordnet is already up-to-date!
[nltk_data] Downloading package punkt to /root/nltk_data...
[nltk_data] Package punkt is already up-to-date!
top5 = ["hotel","room","rooms","hotels"]
for x in top5:
data["Review"] = data["Review"].astype(str).str.replace(x,"")
data.head(2)
data2=data.copy()
def removing_stop_words(txt):
stop_words = set(stopwords.words('english'))
word_tokens = word_tokenize(txt)
filtered_sentence = [w for w in word_tokens if not w in stop_words]
return filtered_sentence
data2["Review"] = data2["Review"].apply(removing_stop_words)
# Making a function to lemmatize
lemmatizer = WordNetLemmatizer()
def lemmatize(data):
lema_data=[]
for j in data:
x=j.lower()
x=lemmatizer.lemmatize(j,pos='n')
x=lemmatizer.lemmatize(j,pos='v')
x=lemmatizer.lemmatize(j,pos='a')
x=lemmatizer.lemmatize(j,pos='r')
x=lemmatizer.lemmatize(x)
lema_data.append(x)
return lema_data
data2["Review"] = data2["Review"].apply(lemmatize)
data2["Review"] = data2["Review"].apply(lambda x:" ".join(token for token in x))
data2.head(2)
X = data2["Review"].values
tokenizer = Tokenizer()
tokenizer.fit_on_texts(X)
fig, ax = plt.subplots()
sns.set(rc={'figure.figsize':(12,9)})
length_dist = [len(x.split(" ")) for x in X]
sns.histplot(length_dist,palette="pastel")
ax.patch.set_visible(False)
ax.set_xlim(0,600)
ax.set_ylim(0,1200)
ax.set_title("Sentence length distribution",fontsize=20);
plt.show()
X = tokenizer.texts_to_sequences(X)
max_length = max([len(x) for x in X])
vocab_size = len(tokenizer.word_index)+1
print("Vocabulary size: {}".format(vocab_size))
print("max length of sentence: {}".format(max_length))
Vocabulary size: 48977
max length of sentence: 1931
# Padding the reviews [Pads sequences to the same length.]
X = pad_sequences(X, padding='post', maxlen=600)
labels = ['1', '2', '3', '4', '5']
y = data['Rating']
y = y.map({1: 0,
2: 1,
3: 2,
4: 3,
5: 4
})
X_train, X_test, y_train, y_test = train_test_split( X, y, test_size=0.1, random_state=42)
EPOCHS
embedding_dim
BATCH_SIZE
units
val_split
model = tf.keras.Sequential([
L.Embedding(vocab_size, int(embedding_dim), input_length=X.shape[1]),
L.Bidirectional(L.LSTM(int(units),return_sequences=True)),
L.Conv1D(64,3),
L.MaxPool1D(),
L.Flatten(),
L.Dropout(0.2),
L.Dense(128, activation="relu"),
L.Dropout(0.2),
L.Dense(64, activation="relu"),
L.Dropout(0.2),
L.Dense(5, activation="softmax")
])
plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)
model.compile(loss=SparseCategoricalCrossentropy(),
optimizer='adam',metrics=['accuracy']
)
history = model.fit(X_train, y_train, epochs=int(EPOCHS), validation_split=float(val_split), batch_size=int(BATCH_SIZE), verbose=2)
Epoch 1/2
130/130 - 427s - loss: 1.1490 - accuracy: 0.4879 - val_loss: 0.8895 - val_accuracy: 0.6078
Epoch 2/2
130/130 - 411s - loss: 0.8167 - accuracy: 0.6324 - val_loss: 0.8481 - val_accuracy: 0.6150
pred = model.predict(X_test)
pred_final = np.argmax(pred,axis=-1)
pred_final
from sklearn.metrics import accuracy_score
print('Accuracy: {}%'.format(round(accuracy_score(pred_final, y_test)*100),2))
Accuracy: 64%
from sklearn.metrics import mean_squared_error
print("Root mean square error: {}".format(round(np.sqrt(mean_squared_error(pred_final,y_test)),3)))
Root mean square error: 0.697
precision recall f1-score support
1 0.74 0.62 0.67 139
2 0.45 0.39 0.42 170
3 0.43 0.42 0.42 216
4 0.53 0.62 0.57 614
5 0.80 0.76 0.78 911
accuracy 0.64 2050
macro avg 0.59 0.56 0.57 2050
weighted avg 0.65 0.64 0.64 2050
model.save("Tripadvisor_BiLSTM.h5")
new_model = tf.keras.models.load_model('Tripadvisor_BiLSTM.h5')
# Check its architecture
new_model.summary()
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
embedding (Embedding) (None, 600, 16) 783632
_________________________________________________________________
bidirectional (Bidirectional (None, 600, 152) 56544
_________________________________________________________________
conv1d (Conv1D) (None, 598, 64) 29248
_________________________________________________________________
max_pooling1d (MaxPooling1D) (None, 299, 64) 0
_________________________________________________________________
flatten (Flatten) (None, 19136) 0
_________________________________________________________________
dropout (Dropout) (None, 19136) 0
_________________________________________________________________
dense (Dense) (None, 128) 2449536
_________________________________________________________________
dropout_1 (Dropout) (None, 128) 0
_________________________________________________________________
dense_1 (Dense) (None, 64) 8256
_________________________________________________________________
dropout_2 (Dropout) (None, 64) 0
_________________________________________________________________
dense_2 (Dense) (None, 5) 325
=================================================================
Total params: 3,327,541
Trainable params: 3,327,541
Non-trainable params: 0
_________________________________________________________________