import streamlit as st
import pickle
import numpy as np
import re
from gensim.models import Word2Vec
from nltk.corpus import stopwords
import nltk
nltk.download('stopwords')

# ===============================
# Word2VecVectorizer class (must be defined before loading pickle)
# ===============================
class Word2VecVectorizer:
    def __init__(self, model):
        self.model = model
        self.size = model.vector_size

    def fit(self, X, y=None):
        return self

    def transform(self, X):
        vectors = []
        for text in X:
            tokens = text.split()
            word_vecs = []
            for w in tokens:
                if w in self.model.wv:
                    vec = self.model.wv[w]
                    # normalize each vector
                    vec = vec / np.linalg.norm(vec)
                    word_vecs.append(vec)
            if word_vecs:
                vectors.append(np.mean(word_vecs, axis=0))
            else:
                vectors.append(np.zeros(self.size))
        return np.array(vectors)

# ===============================
# Streamlit App Config
# ===============================
st.set_page_config(page_title="🎯 Word2Vec Sentiment Classifier")

st.title("🎯 Live Sentiment Classifier (Word2Vec + Logistic Regression)")
#st.write("This app predicts the sentiment of text using a Word2Vec-based Logistic Regression model trained on your dataset.")

# Stopwords
stop_words = set(stopwords.words('english'))

# ===============================
# Load Trained Pipeline
# ===============================
with open("text_classification_word2vec.pkl", "rb") as f:
    pipeline = pickle.load(f)

# ===============================
# User Input
# ===============================
user_input = st.text_area("**Enter a sentence or YouTube comment to analyze:**")

# ===============================
# Prediction
# ===============================
if st.button("Predict Sentiment"):
    if user_input.strip():
        prediction = pipeline.predict([user_input])[0]

        st.markdown(f"## ✅ Prediction: `{prediction.capitalize()}`")

        if prediction.lower() == "positive":
            st.success("😊 Positive Sentiment")
        elif prediction.lower() == "negative":
            st.error("😠 Negative Sentiment")
        else:
            st.info("😐 Neutral Sentiment")
    else:
        st.warning("⚠️ Please enter some text to analyze.")