import streamlit as st
import pandas as pd
from keras.models import load_model
import numpy as np
import re
from Sastrawi.Stemmer.StemmerFactory import StemmerFactory
import nltk
from nltk.tokenize import word_tokenize
import tensorflow as tf

st.set_page_config(
    page_title="Indonesian Sentiment Analysis")
stw = pd.read_csv('https://raw.githubusercontent.com/rizalespe/Dataset-Sentimen-Analisis-Bahasa-Indonesia/master/stopword_tweet_pilkada_DKI_2017.csv')
stpwds_id = stw.values.tolist()
stpwds_id.append(['ada'])
stemmer = StemmerFactory().create_stemmer()
model = load_model('best_model.H5')
nltk.download('punkt')


def run():
    st.title("Indonesian Sentiment Analysis App")
    st.image('https://i.pinimg.com/originals/52/ad/6a/52ad6a11c1dcb1692ff9e321bd520167.gif')
    st.subheader("Enter text to analyze sentiment")

    user_input = st.text_area("Input Text", "")
    if st.button("Analyze"):
        # Case folding
        text1 = user_input.lower()
        # Mention removal
        text2 = re.sub("@[A-Za-z0-9_]+", " ", text1)
        # Hashtags removal
        text3 = re.sub("#[A-Za-z0-9_]+", " ", text2)
        # Newline removal (\n)
        text4 = re.sub(r"\\n", " ",text3)
        # Whitespace removal
        text5 = text4.strip()
        # URL removal
        text6 = re.sub(r"http\S+", " ", text5)
        text7 = re.sub(r"www.\S+", " ", text6)
        # Non-letter removal (such as emoticon, symbol (like μ, $, 兀), etc
        text8 = re.sub("[^A-Za-z\s']", " ", text7)
        # Tokenization
        tokens = word_tokenize(text8)
        # Stopwords removal
        # tokens = [word for word in tokens if word not in stpwds_id]
        tokens1 = []
        for word in tokens:
                    if word not in stpwds_id:
                        tokens1.append(word)
        # Stemming
        #   tokens2 = [stemmer.stem(word) for word in tokens]
        tokens2 = []
        for word in tokens1:
            tokens2.append(stemmer.stem(word))
        # Combining Tokens
        text8 = ' '.join(tokens2)
        
        pred = model.predict([[text8]])
        if pred > 0.3:
            sentiment = "Positive"
        else:
             sentiment = "Negative"
        st.write(f"Sentiment: {sentiment}")

if __name__ == "__main__":
    run()