app.py

import gradio as gr
import pandas as pd
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.cluster import KMeans
import re
from io import BytesIO
import tempfile
from wordcloud import WordCloud, STOPWORDS
import matplotlib.pyplot as plt
import plotly.express as px
from PIL import Image

categories_keywords = {
    "Application Status": ["application status", "application", "status", "submitted", "processing", "pending", "approval", "rejected", "accepted"],
    "Volunteering": ["volunteer", "volunteering", "help out", "assist", "volunteer work", "volunteer opportunities"],
    "Certificates": ["certificate", "certificates", "completion", "certification", "accreditation", "proof", "document", "certified"],
    "Job Opportunities": ["job", "opportunity", "career", "vacancy", "position", "employment", "hiring", "recruitment", "internship"],
    "Surveys and Forms": ["survey", "form", "forms", "questionnaire", "feedback form", "response", "fill out", "submission"],
    "General Queries": ["hello", "hi", "help", "general", "query", "question", "info", "information", "inquiry", "ask"],
    "Spam": ["spam", "unsubscribe", "remove", "stop", "junk", "block", "opt-out"],
    "Rescheduling and Postponing": ["reschedule", "postpone", "delay", "change date", "new time", "rearrange", "shift", "adjust timing"],
    "Contact and Communication Issues": ["contact", "communicate", "communication", "reach out", "phone", "email", "address", "details"],
    "Email and Credentials Issues": ["email", "credentials", "login", "password", "access", "username", "account", "verification", "reset"],
    "Timing and Scheduling": ["timing", "schedule", "scheduling", "time", "appointment", "availability", "calendar", "book", "slot"],
    "Salary and Benefits": ["salary", "benefits", "pay", "compensation", "wages", "earnings", "package", "remuneration", "incentives"],
    "Technical Issues": ["technical", "issue", "problem", "error", "bug", "glitch", "fix", "troubleshoot", "support"],
    "End of Conversation": ["bye", "thank you", "thanks", "goodbye", "see you", "later", "end", "close", "sign off"],
    "Start of Conversation": ["start", "begin", "hello", "hi", "initiate", "greet", "greeting", "open", "commence"],
    "Feedback": ["feedback", "comments", "review", "opinion", "suggestion", "critique", "rating"],
    "Event Inquiries": ["event", "webinar", "meeting", "conference", "session", "seminar", "workshop", "invitation"],
    "Payment Issues": ["payment", "billing", "transaction", "charge", "fee", "invoice", "refund", "receipt"],
    "Registration Issues": ["registration", "register", "sign up", "enroll", "join", "signup", "enrollment"],
    "Service Requests": ["service", "support", "request", "assistance", "help", "aid", "maintenance"],
    "Account Issues": ["account", "profile", "update", "activation", "deactivation", "credentials", "reset"],
    "Product Information": ["product", "service", "details", "info", "information", "specifications", "features"],
    "Order Status": ["order", "status", "tracking", "shipment", "delivery", "purchase", "dispatch"],
    "Miscellaneous": []
}


def categorize_question(question):
    for category, keywords in categories_keywords.items():
        for keyword in keywords:
            if keyword.lower() in question.lower():
                # Check if the question is one word and belongs to 'End of Conversation'
                if category == 'End of Conversation':
                    return category
                # If not 'End of Conversation', return the matched category
                if category != 'End of Conversation':
                    return category
    return 'Miscellaneous'



def preprocess_data(df):
    df.rename(columns={'Question Asked': 'texts'}, inplace=True)
    df['texts'] = df['texts'].astype(str).str.lower()
    df['texts'] = df['texts'].apply(lambda text: re.sub(r'https?://\S+|www\.\S+', '', text))

    def remove_emoji(string):
        emoji_pattern = re.compile("["
                               u"\U0001F600-\U0001F64F"
                               u"\U0001F300-\U0001F5FF"
                               u"\U0001F680-\U0001F6FF"
                               u"\U0001F1E0-\U0001F1FF"
                               u"\U00002702-\U000027B0"
                               u"\U000024C2-\U0001F251"
                               "]+", flags=re.UNICODE)
        return emoji_pattern.sub(r'', string)

    df['texts'] = df['texts'].apply(remove_emoji)

    custom_synonyms = {
        'application': ['form'],
        'apply': ['fill', 'applied'],
        'work': ['job'],
        'salary': ['stipend', 'pay', 'payment', 'paid'],
        'test': ['online test', 'amcat test', 'exam', 'assessment'],
        'pass': ['clear', 'selected', 'pass or not'],
        'result': ['outcome', 'mark', 'marks'],
        'thanks': ["thanks a lot to you", "thankyou so much", "thank you so much", "tysm", "thank you",
                   "okaythank", "thx", "ty", "thankyou", "thank", "thank u"],
        'interview': ["pi"]
    }

    for original_word, synonym_list in custom_synonyms.items():
        for synonym in synonym_list:
            pattern = r"\b" + synonym + r"\b"
            df['texts'] = df['texts'].str.replace(pattern, original_word, regex=True)

    spam_list = ["click here", "free", "recharge", "limited", "discount", "money back guarantee", "aaj", "kal", "mein",
                 "how can i help you", "how can we help you", "how we can help you", "follow", "king", "contacting", "gar",
                 "kirke", "subscribe", "youtube", "jio", "insta", "make money", "b2b", "sent using truecaller"]

    for spam_phrase in spam_list:
        pattern = r"\b" + re.escape(spam_phrase) + r"\b"
        df = df[~df['texts'].str.contains(pattern)]

    def remove_punctuations(text):
        return re.sub(r'[^\w\s]', '', text)

    df['texts'] = df['texts'].apply(remove_punctuations)
    df['texts'] = df['texts'].str.strip()
    df = df[df['texts'] != '']

    # Categorize the texts
    df['Category'] = df['texts'].apply(categorize_question)

    return df

def cluster_data(df, num_clusters):
    vectorizer = TfidfVectorizer(stop_words='english')
    X = vectorizer.fit_transform(df['texts'])

    kmeans = KMeans(n_clusters=num_clusters, random_state=0)
    kmeans.fit(X)
    df['Cluster'] = kmeans.labels_

    return df, kmeans

def generate_wordcloud(df):
    text = " ".join(df['texts'].tolist())
    stopwords = set(STOPWORDS)
    wordcloud = WordCloud(
        width=800,
        height=400,
        background_color='white',
        max_words=300,
        collocations=False,
        min_font_size=10,
        max_font_size=200,
        stopwords=stopwords,
        prefer_horizontal=1.0,
        scale=2,
        relative_scaling=0.5,
        random_state=42
    ).generate(text)
    
    plt.figure(figsize=(15, 7))
    plt.imshow(wordcloud, interpolation='bilinear')
    plt.axis('off')
    buf = BytesIO()
    plt.savefig(buf, format='png')
    buf.seek(0)
    img = Image.open(buf)
    return img

def generate_bar_chart(df, num_clusters_to_display):
    # Exclude common words from the top words
    common_words = {'i', 'you', 'thanks', 'thank', 'ok', 'okay', 'sure', 'done'}
    
    top_categories = df['Category'].value_counts().index[:num_clusters_to_display]
    df_top_categories = df[df['Category'].isin(top_categories)]
    
    category_top_words = df_top_categories.groupby('Category', observed=False)['texts'].apply(lambda x: ' '.join(x)).reset_index()
    category_top_words['top_word'] = category_top_words['texts'].apply(lambda x: ' '.join([word for word in pd.Series(x.split()).value_counts().index if word not in common_words][:3]))
    category_sizes = df_top_categories['Category'].value_counts().reset_index()
    category_sizes.columns = ['Category', 'Count']
    category_sizes = category_sizes.merge(category_top_words[['Category', 'top_word']], on='Category')
    
    fig = px.bar(category_sizes, x='Category', y='Count', text='top_word', title='Category Frequency with Top Words')
    fig.update_traces(textposition='outside')
    fig.update_layout(xaxis_title='Category', yaxis_title='Frequency', showlegend=False)
    
    buf = BytesIO()
    fig.write_image(buf, format='png')
    buf.seek(0)
    img = Image.open(buf)
    return img

def main(file, num_clusters_to_display):
    try:
        df = pd.read_csv(file)
        
        # Filter by 'Fallback Message shown'
        df = df[df['Answer'] == 'Fallback Message shown']
        
        df = preprocess_data(df)

        df = df[df['Category'] != 'Miscellaneous']
        
        # Get category sizes and sort by size in ascending order
        category_sizes = df['Category'].value_counts().reset_index()
        category_sizes.columns = ['Category', 'Count']
        sorted_categories = category_sizes.sort_values(by='Count', ascending=False)['Category'].tolist()
        sorted_categories_sm = category_sizes.sort_values(by='Count', ascending=True)['Category'].tolist()
        
        # Get the largest x categories as specified by num_clusters_to_display
        largest_categories = sorted_categories[:num_clusters_to_display]
        smallest_categories = sorted_categories_sm[:num_clusters_to_display]
        
        # Filter the dataframe to include only the largest categories
        filtered_df = df[df['Category'].isin(largest_categories)]
        filtered_cloud_df = df[df['Category'].isin(smallest_categories)]
        
        # Sort the dataframe by Category
        filtered_df = filtered_df.sort_values(by='Category')
        filtered_cloud_df = filtered_cloud_df.sort_values(by='Category')
        
        wordcloud_img = generate_wordcloud(filtered_cloud_df)
        bar_chart_img = generate_bar_chart(df, num_clusters_to_display)

        with tempfile.NamedTemporaryFile(delete=False, suffix=".csv") as tmpfile:
            filtered_df.to_csv(tmpfile.name, index=False)
            csv_file_path = tmpfile.name

        return csv_file_path, wordcloud_img, bar_chart_img
    except Exception as e:
        print(f"Error: {e}")
        return str(e), None, None

interface = gr.Interface(
    fn=main,
    inputs=[
        gr.File(label="Upload CSV File (.csv)"),
        gr.Slider(label="Number of Categories to Display", minimum=1, maximum=15, step=1, value=5)
    ],
    outputs=[
        gr.File(label="Categorized Data CSV"),
        gr.Image(label="Word Cloud"),
        gr.Image(label="Bar Chart")
    ],
    title="Unanswered User Queries Categorization",
    description="Categorize unanswered user queries into predefined categories"
)

interface.launch(share=True)