import streamlit as st
import tensorflow as tf
from tensorflow.keras.datasets import imdb
from tensorflow.keras.preprocessing import sequence
from tensorflow.keras.preprocessing.sequence import pad_sequences
import numpy as np

word_index = imdb.get_word_index()
max_num_palabras = 2000

def reviewnueva(review, word_index, max_num_palabras):
  sequence = []
  for word in review.split():
    index = word_index.get(word.lower(), 0)
    if index < max_num_palabras:
      sequence.append(index)
  return sequence

model = tf.keras.models.load_model("opiniones.h5")

def predict_sentimiento(review):
    sequence =  reviewnueva(review, word_index)
    prediccion = model.predict(sequence)
    if prediccion[0][0]>=0.5:
        sentimiento = "Positivo"
    else:
        sentimiento = "Negativo"
    return sentimiento

st.title("Ingrese una review para poder calificar como positiva o negativa")
review = st.text_area("Ingrese reseña aquí", height = 200)
if st.button("Predecir sentimiento"):
    if review:
        sentimiento = predict_sentimiento(review)
        st.write(f'El sentimiento es: {sentimiento}')
    else:
        st.write(f'Ingrese una review')