unlabelledmtb.py

import pandas as pd
from text_preprocessor import preprocess_text
from textblob import TextBlob
from datetime import datetime
import csv  # Importing csv for logging
from prettytable import PrettyTable
import matplotlib.pyplot as plt
import seaborn as sns
import nltk
nltk.download('punkt_tab')


# Load the dataset
df = pd.read_csv('ytcomments100list.csv')  # Assuming your updated CSV has the correct column names

# Preprocess the comments
df['cleaned_comment'] = df['comment'].apply(preprocess_text)

# Function to calculate polarity of the comment
def get_polarity(text):
    """
    Calculate polarity of a comment using TextBlob.
    Polarity ranges from -1 (negative) to 1 (positive).
    """
    analysis = TextBlob(text)
    return analysis.sentiment.polarity

# Apply polarity calculation
df['polarity'] = df['cleaned_comment'].apply(get_polarity)

# Determine if a comment is negative
def is_negative(polarity):
    """
    Classify a comment as negative if polarity < 0.
    """
    return polarity < 0

# Add a column indicating negativity
df['is_negative'] = df['polarity'].apply(is_negative)

# Count negative comments
negative_count = df['is_negative'].sum()

# Calculate threshold (e.g., 40% of total comments)
threshold = 0.4 * len(df)
flagged = negative_count > threshold

# Print results to console
print(f"\nTotal Comments: {len(df)}")
print(f"Negative Comments: {negative_count}")
print(f"Threshold: {threshold}\n\n")
if flagged:
    print("Warning: This may be a case of cyberbullying!")
else:
    print("No significant negative activity detected.")

# Function to log messages to a CSV file
def log_message_csv(message):
    timestamp = datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    
    # Check if the CSV file exists; if not, create it and write headers
    try:
        with open('log.csv', 'x', newline='') as file:  # 'x' mode creates the file if it doesn't exist
            writer = csv.writer(file)
            writer.writerow(['timestamp', 'message'])  # Write headers
    except FileExistsError:
        pass  # If the file exists, no need to write headers again

    with open('log.csv', 'a', newline='') as file:
        writer = csv.writer(file)
        writer.writerow([timestamp, message])

# Log the message based on the flag
if flagged:
    log_message_csv(f"Warning: Cyberbullying detected. {negative_count} negative comments out of {len(df)}.")
else:
    log_message_csv(f"No significant negative activity detected. {negative_count} negative comments out of {len(df)}.")

# Save the results to a new CSV file (optional, for further analysis)
df.to_csv('processed_comments.csv', index=False)

# PrettyTable function to display a summary of the comments and their polarity
'''def display_summary(df):
    table = PrettyTable()
    table.field_names = ["Comment", "Polarity", "Negative"]

    for index, row in df.iterrows():
        table.add_row([row['comment'], row['polarity'], row['is_negative']])

    print(table)'''

# Plotting the sentiment distribution (Negative vs Positive Comments)
def plot_sentiment_distribution(df):
    sentiment_counts = df['is_negative'].value_counts()
    sentiment_counts = sentiment_counts.sort_index()  # Ensure consistent ordering of positive and negative

    plt.figure(figsize=(6, 6))
    sns.barplot(
        x=sentiment_counts.index, 
        y=sentiment_counts.values, 
        hue=sentiment_counts.index,  # Assign x to hue
        palette='Blues', 
        dodge=False  # Optional, avoids separate bars for hue
    )
    plt.title('Sentiment Distribution of Comments')
    plt.xlabel('Sentiment (Negative/Positive)')
    plt.ylabel('Count')
    plt.xticks([0, 1], ['Positive', 'Negative'], rotation=45)
    plt.legend([], [], frameon=False)  # Disable legend
    plt.show()


# Call the functions to display the summary and plot the distribution
# display_summary(df)
plot_sentiment_distribution(df)