app.py

import streamlit as st
from transformers import pipeline

# Set page title and icon
st.set_page_config(page_title="Privacy Policy Relation Extraction", page_icon="🐢")


@st.cache_data
def get_results(text):
    pipe = pipeline(
        "text-classification",
        model="PaDaS-Lab/privacy-policy-relation-extraction",
        return_all_scores=True,
        framework="pt",
    )
    return pipe(text)


st.title("Privacy Policy Relation Extraction")

example = st.text_area(
    "Enter text:",
    value="We store your basic account information, including your name, username, and email address until you ask us to delete them.",
    height=150,
)

if st.button("Analyze"):
    with st.spinner("Processing..."):
        results = get_results(example)
        st.session_state.results = results

if "results" in st.session_state:
    threshold = st.slider(
        "Confidence threshold:", min_value=0.0, max_value=1.0, value=0.5
    )
    filtered_results = [
        result for result in st.session_state.results[0] if result["score"] >= threshold
    ]

    sorted_results = sorted(filtered_results, key=lambda x: x["score"], reverse=True)

    if sorted_results:
        for result in sorted_results:
            cols = st.columns([3, 5, 0.5])
            with cols[0]:
                st.write(f"**{result['label']}**")
            with cols[1]:
                st.progress(int(result["score"] * 100))
            with cols[2]:
                st.write(f"**{result['score']:.2f}**")
    else:
        st.warning("No relations found with the specified threshold.")

st.markdown(
    """
<style>
    .reportview-container {
        background: #f0f2f6;
        padding: 20px;
    }
    .stButton button {
        border: 0.5;
        transition: background-color 0.3s, transform 0.2s;
        border-radius: 10px;
        box-shadow: 0 1px 3px rgba(0, 0, 0, 0.08);
    }
    .stButton button:hover {
        transform: translateY(-1px);
    }
</style>
""",
    unsafe_allow_html=True,
)