app.py

import streamlit as st
import pandas as pd
import numpy as np
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.neighbors import NearestNeighbors
import matplotlib.pyplot as plt
import seaborn as sns
from collections import Counter

st.set_page_config(page_title="🎵 Lyrics-Based Song Recommendations")

# Load dataset
@st.cache_data
def load_data():
    df = pd.read_csv("spotify_millsongdata.csv")  # Update with actual file path
    df = df.dropna(subset=["text"])  # Remove missing lyrics
    return df

df = load_data()

# Convert lyrics into numerical features using TF-IDF
vectorizer = TfidfVectorizer(stop_words="english", max_features=5000)
lyrics_matrix = vectorizer.fit_transform(df["text"])

# Train KNN Model
knn = NearestNeighbors(n_neighbors=5, metric="cosine")
knn.fit(lyrics_matrix)

# Streamlit UI
st.title("🎶 Lyrics-Based Song Recommendation System")

st.markdown(
    "Discover songs that match your favorite lyrics! This app uses **TF-IDF** and **KNN** to find songs with similar lyrical content."
)

# Tabs for better UI
tab1, tab2 = st.tabs(["📊 Dataset Overview", "🎤 Lyrics-Based Recommendation"])

with tab1:
    # Dataset Sample
    sample_df = df.sample(20)
    st.dataframe(sample_df[["song", "artist", "text"]])

    # Expander for Dataset Statistics
    with st.expander("📊 Dataset Statistics"):
        # Dataset Statistics
        total_songs = df.shape[0]
        unique_artists = df["artist"].nunique()
        avg_lyrics_length = df["text"].apply(lambda x: len(x.split())).mean()

        st.write(f"📊 **Total Songs**: {total_songs}")
        st.write(f"🎤 **Unique Artists**: {unique_artists}")
        st.write(f"📖 **Average Lyrics Length**: {avg_lyrics_length:.2f} words")

    # Expander for Lyrics Length Distribution
    with st.expander("📖 Lyrics Length Distribution (Word Count per Song)"):
        # Lyrics Length Distribution
        lyrics_length = df["text"].apply(lambda x: len(x.split()))
        fig, ax = plt.subplots(figsize=(8, 4))
        sns.histplot(lyrics_length, kde=True, ax=ax, color="skyblue")
        ax.set_xlabel("Word Count")
        ax.set_ylabel("Number of Songs")
        st.pyplot(fig)

    # Most Frequent Artists
    with st.expander("🎤 Most Frequent Artists in the Dataset"):
        artist_counts = df["artist"].value_counts().head(10)
        fig, ax = plt.subplots(figsize=(8, 4))
        sns.barplot(y=artist_counts.index, x=artist_counts.values, ax=ax, palette="mako")
        ax.set_xlabel("Number of Songs")
        ax.set_ylabel("Artist")
        st.pyplot(fig)

with tab2:
    st.subheader("Enter Lyrics Snippet")
    user_lyrics = st.text_area("Type lyrics snippet:", "")
    
    if st.button("🎶 Get Recommendations") and user_lyrics.strip():
        # Convert user input into the same TF-IDF space
        user_vector = vectorizer.transform([user_lyrics])
        
        # Find similar songs
        distances, indices = knn.kneighbors(user_vector)
        
        st.subheader("Recommended Songs:")
        recommendations = []
        for i, idx in enumerate(indices[0]):  
            recommended_song = df.iloc[idx]["song"]
            recommended_artist = df.iloc[idx]["artist"]
            similarity_score = 1 - distances[0][i]  # Convert cosine distance to similarity
            recommendations.append((recommended_song, recommended_artist, similarity_score))
            st.write(f"🎶 **{recommended_song}** - {recommended_artist}  (Similarity: `{similarity_score:.2f}`)")