app.py

import streamlit as st
from keybert import KeyBERT

# Create a KeyBERT instance
kw_model = KeyBERT()

# Define the Streamlit app
def main():
    st.title("Keyword Extraction")
    st.write("Enter your document below:")

    # Get user input
    doc = st.text_area("Document")

    # Get user choice for stopwords removal (default checkbox)
    remove_stopwords = st.checkbox("Remove Stopwords", value=True)

    # Get user choice for MMR (default checkbox)
    apply_mmr = st.checkbox("Apply Maximal Marginal Relevance (MMR)", value=True)

    # Get user choice for number of results (slider)
    num_results = st.slider("Number of Results", min_value=1, max_value=30, value=5, step=1)

    # Extract keywords
    if st.button("Extract Keywords"):
        keywords = kw_model.extract_keywords(doc, stop_words=None if remove_stopwords else "english")

        if apply_mmr:
            # Apply Maximal Marginal Relevance (MMR)
            selected_keywords = []
            selected_keywords.append(keywords[0])  # Select the top-scoring keyword

            # Set the MMR hyperparameters
            lambda_param = 0.7  # Weight for the trade-off between relevance and diversity

            for i in range(1, num_results):
                selected_keywords_scores = [kw[1] for kw in selected_keywords]
                remaining_keywords = [kw for kw in keywords if kw[0] not in [kw[0] for kw in selected_keywords]]
                mmr_scores = kw_model.maximal_marginal_relevance(doc, remaining_keywords, selected_keywords_scores, lambda_param)
                max_mmr_index = mmr_scores.index(max(mmr_scores))
                selected_keywords.append(remaining_keywords[max_mmr_index])

            keywords = selected_keywords  # Update keywords with MMR-selected keywords

        st.write(f"Top {num_results} Keywords:")
        for keyword, score in keywords:
            st.write(f"- {keyword} (Score: {score})")

# Run the app
if __name__ == "__main__":
    main()