# import essential libraraies
import streamlit as st
import pandas as pd
from transformers import pipeline #for pre-trained model
# for graphing
import plotly.express as px
import matplotlib.pyplot as plt
import plotly.graph_objs as go
from wordcloud import WordCloud
# Load the pre-trained sentiment analysis model
sentiment_model = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")
# Define user interface
# Set page title
st.set_page_config(page_title="Hotel Reviews Sentiment", page_icon=":hotel:",layout='wide')
# Add a header
st.image("Header.png", use_column_width=True)
# Define the format of the file to be uploaded
st.write("
", unsafe_allow_html=True)
file_format_link = "https://drive.google.com/file/d/1B6Glpna9kZsakgjpWStfhbxI20EoGsnv/view?usp=sharing"
message = f"⚠️ㅤPlease stick to the given layout when uploading a file. You can download the sample file layout from [here]( {file_format_link})."
st.write(message)
# Uploading the file
file = st.file_uploader("",type=["csv"])
if file is not None:
# Read the CSV file into a Pandas DataFrame
df = pd.read_csv(file)
# Print total number of reviews to analyse
st.markdown(f"Total reviews: {len(df)}
", unsafe_allow_html=True)
st.markdown(
f' Distribution of Reviews
',
unsafe_allow_html=True
)
# Apply the sentiment analysis model
df["Sentiment"] = df["Review"].apply(lambda x: sentiment_model(x)[0]["label"])
# Building the dashboard
# Generate pie chart
colors = ['#30C3C4', '#D1DDDE']
sentiment_counts = df["Sentiment"].value_counts()
fig = px.pie(sentiment_counts, values=sentiment_counts.values, names=sentiment_counts.index,
color_discrete_sequence=colors)
st.plotly_chart(fig, use_container_width=True)
# Create word clouds for positive and negative reviews
positive_reviews = " ".join(df[df["Sentiment"] == "POSITIVE"]["Review"].tolist())
negative_reviews = " ".join(df[df["Sentiment"] == "NEGATIVE"]["Review"].tolist())
# Diplay wordcloud in two columns
col1, col2 = st.columns(2)
with col1:
st.markdown(
f'Positive Reviews
',
unsafe_allow_html=True
)
wc_pos = WordCloud(width=800, height=600, background_color="white", colormap="winter").generate(positive_reviews)
st.image(wc_pos.to_array(),use_column_width=True)
with col2:
st.markdown(
f'Negative Reviews
',
unsafe_allow_html=True
)
wc_neg = WordCloud(width=800, height=600, background_color="white", colormap="winter").generate(negative_reviews)
st.image(wc_neg.to_array(),use_column_width=True)
# Display the sentiment of each review as a dataframe
st.markdown(
f' Reviews in depth
',
unsafe_allow_html=True
)
# Add a filter for sentiments
filter_sentiment = st.selectbox("", ["ALL", "POSITIVE", "NEGATIVE"])
# Filter the dataframe
if filter_sentiment != "ALL":
df = df[df['Sentiment'] == filter_sentiment]
# Max number of rows to display at a time
max_rows = 10
# Table generation
table_html = (df.style
.set_properties(**{'text-align': 'left'})
.set_table_styles([{'selector': 'th', 'props': [('border', '0px')]},
{'selector': 'td', 'props': [('border', '0px')]}])
.set_table_attributes('style="position: sticky; top: 0;"')
.to_html(index=False, escape=False))
# Scrollable content
st.write(f'{table_html}
', unsafe_allow_html=True,header=True,sticky_header=True)
#save output as csv
def convert_df(df):
# IMPORTANT: Cache the conversion to prevent computation on every rerun
return df.to_csv().encode('utf-8')
csv = convert_df(df)
# Download button
st.write("
", unsafe_allow_html=True)
st.download_button(
label="Download data as CSV",
data=csv,
file_name='Review Sentiments.csv'
)
# Footnote
st.write("
", unsafe_allow_html=True)
st.caption('crafted with ❤️
', unsafe_allow_html=True)