app.py

import gradio as gr
from gnews import GNews
import pandas as pd
from transformers import pipeline
from datetime import datetime, timedelta


def discard_old_rows(df):
    # Convert the 'published date' column to datetime
    df['published date'] = pd.to_datetime(df['published date'], format='%a, %d %b %Y %H:%M:%S %Z')
    
    # Get the current date
    current_date = datetime.utcnow()
    
    # Calculate the date two months ago
    two_months_ago = current_date - timedelta(days=60)
    
    # Filter the DataFrame to keep only the rows with 'published date' within the last two months
    df_filtered = df[df['published date'] >= two_months_ago]
    
    return df_filtered


def extract_and_clean_titles(df):
    # Initialize an empty list to store the cleaned titles
    values_list = []

    # Iterate over each value in the 'title' column of the DataFrame
    for value in df['title']:
        # Find the position of the first hyphen in the title
        index = value.find('-')

        # Extract the part of the title before the hyphen
        # If there's no hyphen, use the entire title
        extracted_value = value[:index] if index >= 0 else value

        # Remove any occurrences of '...' from the extracted value
        cleaned_value = extracted_value.replace('...', '')

        # Append the cleaned value to the list
        values_list.append(cleaned_value)

    # Return the list of cleaned titles
    return values_list


def analyze_sentiments(values_list, sentiment_analysis):
    # Initialize an empty list to store the sentiment predictions
    prediction = []

    # Iterate over each news title in the values_list
    for news in values_list:
        # Perform sentiment analysis on the current news title
        sentiment = sentiment_analysis(news)

        # Append the resulting sentiment to the prediction list
        prediction.append(sentiment)

    # Return the list of sentiment predictions
    return prediction


def calculate_weighted_average(predictions):
    # Initialize the weighted average score to zero
    weighted_avg = 0

    # Iterate over each prediction in the predictions list
    for i in predictions:
        # Check if the label of the first sentiment prediction is 'positive'
        if i[0]['label'] == 'positive':
            # Add the score to the weighted average (positive sentiment)
            weighted_avg += 1 * i[0]['score']
        # Check if the label of the first sentiment prediction is 'negative'
        elif i[0]['label'] == 'negative':
            # Subtract the score from the weighted average (negative sentiment)
            weighted_avg += -1 * i[0]['score']

    # Calculate the weighted average by dividing by the number of predictions
    weighted_avg /= len(predictions)

    # Return the calculated weighted average
    return weighted_avg


def sentiment_pie_chart(predictions):
    """
    Generates a pie chart for sentiment distribution.
    """
    positive_count = 0
    negative_count = 0
    neutral_count = 0

    for item in predictions:
        label = item[0]['label']
        if label == 'positive':
            positive_count += 1
        elif label == 'negative':
            negative_count += 1
        elif label == 'neutral':
            neutral_count += 1

    labels = ['Positive', 'Negative', 'Neutral']
    sizes = [positive_count, negative_count, neutral_count]
    colors = ['#66BB6A', '#EF5350', '#42A5F5']

    fig, ax = plt.subplots()
    ax.pie(sizes, labels=labels, colors=colors, autopct='%1.1f%%', startangle=90, pctdistance=0.85)
    center_circle = plt.Circle((0, 0), 0.70, fc='white')
    fig.gca().add_artist(center_circle)
    ax.axis('equal')
    plt.title('Sentiment Analysis Results')
    return fig


def main(stock):
    #Specifying model 
    model="mrm8488/distilroberta-finetuned-financial-news-sentiment-analysis"
    
    #Scraping top data from google news
    google_news = GNews()
    Company_news=google_news.get_news(stock + "share")
    df=pd.DataFrame(Company_news)

    #Discarding old rows
    df=discard_old_rows(df)

    #Cleaning the titles for sentiment analysis
    values_list=extract_and_clean_titles(df)
    
    #Sentiment Analysis
    sentiment_analysis = pipeline(model=model)

    #Predictions
    predictions=analyze_sentiments(values_list,sentiment_analysis)

    #Weighted Average
    weighted_avg=calculate_weighted_average(predictions)

    #Pie-Chart
    pie_chart = sentiment_pie_chart(predictions)

    return f'Weighted Sentiment Score: {weighted_avg:.2f}', pie_chart

iface = gr.Interface(
    fn=main,
    inputs="textbox",
    outputs=["textbox","image"]
)

if __name__ == "__main__":
    iface.launch()