app.py

import streamlit as st
import pandas as pd
from transformers import T5Tokenizer, T5ForConditionalGeneration, pipeline
from bertopic import BERTopic
import torch

# Initialize ARAT5 model and tokenizer for topic modeling
tokenizer = T5Tokenizer.from_pretrained("UBC-NLP/araT5-base")
model = T5ForConditionalGeneration.from_pretrained("UBC-NLP/araT5-base")

# Emotion classification pipeline for Arabic (use an Arabic emotion classification model)
emotion_classifier = pipeline("text-classification", model="d0r13n/ara-bert-base-arabic-emotion")

# Function to get embeddings from ARAT5 for topic modeling
def generate_embeddings(texts):
    # Tokenize the Arabic text for ARAT5
    inputs = tokenizer(texts, return_tensors="pt", padding=True, truncation=True, max_length=512)
    with torch.no_grad():
        # Use ARAT5 to generate embeddings
        outputs = model.encoder(input_ids=inputs['input_ids'])
    # Extract the embeddings (mean of hidden states for simplicity)
    return outputs.last_hidden_state.mean(dim=1).numpy()

# Function to process the CSV file and return emotion and topic model
def process_file(uploaded_file):
    # Load CSV
    df = pd.read_csv(uploaded_file)
    
    # Display basic info about the CSV
    st.write("CSV Loaded Successfully!")
    st.write(f"Data Preview: {df.head()}")
    
    # Ensure 'date' column is in datetime format and extract the year
    df['date'] = pd.to_datetime(df['date'], errors='coerce')  # Replace 'date' with your actual column name
    df['year'] = df['date'].dt.year
    
    # Modify this to use the 'poem' column that contains the Arabic poems
    texts = df['poem'].dropna().tolist()  # Replace 'poem' with your actual column name
    
    # Emotion Classification: Classify emotions for each poem (Arabic)
    emotions = [emotion_classifier(text)[0]['label'] for text in texts]
    df['emotion'] = emotions
    
    # Topic Modeling using ARAT5 embeddings
    embeddings = generate_embeddings(texts)
    topic_model = BERTopic()
    topics, _ = topic_model.fit_transform(embeddings)
    df['topic'] = topics
    
    # Return the processed dataframe
    return df

# Streamlit App
st.title("Arabic Poem Topic Modeling & Emotion Classification with ARAT5")
st.write("Upload a CSV file to perform topic modeling and emotion classification on Arabic poems.")

# File upload widget
uploaded_file = st.file_uploader("Choose a CSV file", type=["csv"])

# If file is uploaded, process and display results
if uploaded_file is not None:
    result_df = process_file(uploaded_file)

    # Show date selection widgets
    st.write("### Filter by Date Range")
    start_date = st.date_input("Start Date", value=pd.to_datetime(result_df['date'].min()))
    end_date = st.date_input("End Date", value=pd.to_datetime(result_df['date'].max()))
    
    # Filter data based on selected date range
    filtered_df = result_df[(result_df['date'] >= start_date) & (result_df['date'] <= end_date)]
    
    # Display filtered data
    st.write(f"Filtered Data (Poems from {start_date} to {end_date}):")
    st.write(filtered_df[['poet_name', 'era', 'poem', 'emotion', 'topic', 'date']])

    # Create buttons to show different summaries
    summary_type = st.radio("Select Summary Type:", 
                            ("Emotion and Topic Summary by Date Range", 
                             "Global Emotion and Topic Summary"))

    # Display the selected summary
    if summary_type == "Emotion and Topic Summary by Date Range":
        st.write("Emotion and Topic Summary for Selected Date Range:")
        
        # Emotion Distribution in Date Range
        emotion_counts = filtered_df['emotion'].value_counts()
        st.write("Emotion Counts in Date Range:")
        st.write(emotion_counts)
        
        # Topic Distribution in Date Range
        topic_counts = filtered_df['topic'].value_counts()
        st.write("Topic Counts in Date Range:")
        st.write(topic_counts)
        
        # Visualize emotion distribution over the selected range (optional)
        st.bar_chart(emotion_counts, use_container_width=True)
        
        # Visualize topic distribution over the selected range (optional)
        st.bar_chart(topic_counts, use_container_width=True)

    elif summary_type == "Global Emotion and Topic Summary":
        st.write("Global Emotion and Topic Summary (All Poems):")
        global_emotion_count = result_df['emotion'].value_counts().to_dict()
        global_topic_count = result_df['topic'].value_counts().to_dict()
        
        st.write(f"Emotion Distribution: {global_emotion_count}")
        st.write(f"Topic Distribution: {global_topic_count}")