app.py

import streamlit as st

# Set up the Streamlit page - this must be the first st command
st.set_page_config(
    page_title="Paper Classification Service",
    page_icon="📚",
    layout="wide"
)

import PyPDF2
import io
from model import PaperClassifier

# Initialize the classifier with model selection
@st.cache_resource
def load_classifier(model_type):
    return PaperClassifier(model_type)

# Cache the PDF text extraction
@st.cache_data
def extract_pdf_text(pdf_bytes):
    """Extract text from PDF and try to separate title and abstract"""
    pdf_reader = PyPDF2.PdfReader(io.BytesIO(pdf_bytes))
    text = ""
    for page in pdf_reader.pages:
        text += page.extract_text() + "\n"
    
    # Try to extract title and abstract
    lines = text.split('\n')
    title = lines[0] if lines else ""
    abstract = "\n".join(lines[1:]) if len(lines) > 1 else ""
    
    return title.strip(), abstract.strip()

# Get available models for selection
available_models = list(PaperClassifier.AVAILABLE_MODELS.keys())

# Add model selection to sidebar
st.sidebar.title("Model Settings")
selected_model = st.sidebar.selectbox(
    "Select Model",
    available_models,
    index=0,
    help="Choose the model to use for classification"
)

# Display model information
model_info = PaperClassifier.AVAILABLE_MODELS[selected_model]
st.sidebar.markdown(f"""
### Selected Model
**Name:** {model_info['name']}  
**Description:** {model_info['description']}
""")

# Initialize the classifier with selected model
classifier = load_classifier(selected_model)

# Title and description
st.title("📚 Academic Paper Classification")
st.markdown("""
This service helps you classify academic papers into different categories.
You can either:
- Enter the paper's title and abstract separately
- Upload a PDF file
""")

# Create two columns for input methods
col1, col2 = st.columns(2)

with col1:
    st.subheader("Option 1: Manual Input")
    
    # Title input
    title_input = st.text_input(
        "Paper Title:",
        placeholder="Enter the paper title..."
    )
    
    # Abstract input
    abstract_input = st.text_area(
        "Paper Abstract (optional):",
        height=200,
        placeholder="Enter the paper abstract (optional)..."
    )
    
    if st.button("Classify Paper"):
        if title_input.strip():
            with st.spinner("Classifying..."):
                result = classifier.classify_paper(
                    title=title_input,
                    abstract=abstract_input if abstract_input.strip() else None
                )
                
                st.success("Classification Complete!")
                st.write(f"**Input Type:** {result['input_type'].replace('_', ' ').title()}")
                st.write(f"**Model Used:** {result['model_used']}")
                
                # Show top categories
                st.subheader("Top Categories (95% Confidence)")
                total_prob = 0
                for cat_info in result['top_categories']:
                    prob = cat_info['probability']
                    total_prob += prob
                    st.progress(prob, text=f"{cat_info['category']} ({cat_info['arxiv_category']}): {prob:.1%}")
                
                st.info(f"Total probability of shown categories: {total_prob:.1%}")
        else:
            st.warning("Please enter at least the paper title.")

with col2:
    st.subheader("Option 2: PDF Upload")
    uploaded_file = st.file_uploader("Upload a PDF file", type="pdf")
    
    if uploaded_file is not None:
        if st.button("Classify PDF"):
            try:
                with st.spinner("Processing PDF..."):
                    # Extract title and abstract from PDF
                    title, abstract = extract_pdf_text(uploaded_file.read())
                    
                    if not title:
                        st.error("Could not extract title from PDF.")
                        st.stop()
                    
                    # Show extracted text
                    with st.expander("Show extracted text"):
                        st.write("**Extracted Title:**")
                        st.write(title)
                        if abstract:
                            st.write("**Extracted Abstract:**")
                            st.write(abstract)
                    
                    # Classify the paper
                    result = classifier.classify_paper(
                        title=title,
                        abstract=abstract if abstract else None
                    )
                    
                    st.success("Classification Complete!")
                    st.write(f"**Input Type:** {result['input_type'].replace('_', ' ').title()}")
                    st.write(f"**Model Used:** {result['model_used']}")
                    
                    # Show top categories
                    st.subheader("Top Categories (95% Confidence)")
                    total_prob = 0
                    for cat_info in result['top_categories']:
                        prob = cat_info['probability']
                        total_prob += prob
                        st.progress(prob, text=f"{cat_info['category']} ({cat_info['arxiv_category']}): {prob:.1%}")
                    
                    st.info(f"Total probability of shown categories: {total_prob:.1%}")
            except Exception as e:
                st.error(f"Error processing PDF: {str(e)}")

# Add information about the models
st.sidebar.markdown("---")
st.sidebar.title("Available Models")
st.sidebar.markdown("""
- **DistilBERT**: Fast and lightweight
- **DeBERTa v3**: Advanced performance
- **T5**: Versatile text-to-text
- **RoBERTa**: Strong performance
- **SciBERT**: Specialized for science
""")

# Add footer
st.markdown("---")
st.markdown("Made with ❤️ using Streamlit and Transformers")