import streamlit as st
from tensorflow.keras.models import load_model
from sklearn.feature_extraction.text import CountVectorizer
from textblob import TextBlob
from nltk.stem import PorterStemmer
import numpy as np

pr=PorterStemmer()
def lemmafn(text):
    words=TextBlob(text).words
    return [pr.stem(word) for word in words]
vect=CountVectorizer(stop_words="english",ngram_range=(1,3),max_features=10000)

model=load_model("model.h5")

st.title("Predict Comments Toxicity")
comment=st.text_area("Comment")
if comment is not None:
    coment=comment.lower()
    comment=comment.replace("[^\w\s]","",)
    comment=comment.replace("\d+","")
    comment=comment.replace("\n","")
    comment=[comment]
    if st.button("Predict"):
       data=vect.fit_transform(comment)
       prediction=model.predict(data)
       predicted_class=np.argmax(prediction)
       st.write("1st->Toxicity,2nd->Severe Toxicity,3rd->Obscene,4th->Is it a Threat?,5th->Insult,6th->Identity Hate",predicted_class)