import streamlit as st
import pandas as pd
import numpy as ny
from keras.preprocessing.text import Tokenizer
from keras.preprocessing.sequence import pad_sequences
from sklearn.preprocessing import LabelEncoder
from keras.layers import *
from keras import Model

map_id = {
  0: "sadness",
  1: "anger",
  2: "love",
  3: "surprise",
  4: "fear",
  5: "joy"
}
map_emotion = {
  "sadness": 0,
  "anger": 1,
  "love": 2,
  "surprise": 3,
  "fear": 4,
  "joy": 5
}

train = pd.read_csv('/train.csv')
for index, row in train.iterrows():
  row['emotion'] = map_emotion[row['emotion']]
tokenizer = Tokenizer()
tokenizer.fit_on_texts(train.text)
Length = len(tokenizer.word_index) + 1
x_train = pad_sequences(tokenizer.texts_to_sequences(train.text), maxlen=30)
encoder = LabelEncoder()
encoder.fit(train["emotion"].to_list())
y_train = encoder.transform(train["emotion"].to_list())
y_train = y_train.reshape(-1, 1)

embedding_layer = Embedding(Length,
                           64,
                           input_length=30)

input_seq = Input(shape=(x_train.shape[1],))
x = embedding_layer(input_seq)
x = LSTM(10, return_sequences=True) (x)
x = Flatten() (x)
output = Dense(encoder.classes_.shape[0], activation="softmax") (x)

model = Model(input_seq, output)
model.compile(optimizer='adam',
             loss="sparse_categorical_crossentropy",
             metrics=["accuracy"])
model.fit(x_train, y_train, epochs=20, batch_size=32,
         validation_data=(x_val, y_val))
class Predict:
    def __init__(self, model, tokenizer):
        self.model = model
        self.tokenizer = tokenizer
    
    def predict(self, txt):
        x = pad_sequences(self.tokenizer.texts_to_sequences([txt]), maxlen=30)
        x = self.model(x)
        x = ny.argmax(x)
        return map_id[x]
predict = Predict(model, tokenizer)


st.title("TONE DETECTION | BCS WINTER PROJECT")
st.write("Enter a sentence to analyze text's Tone:")

user_input = st.text_input("")
if user_input:
    result = predict.predict(user_input)
    st.write(f"TONE: {result}")