import pandas as pd
import numpy as np
df = pd.read_csv('CrimeVsNoCrimeArticles.csv')
titles = np.array(df['title'].to_list())
labels = np.array(df['is_crime_report'].to_list())
import gradio as gr
import re
import nltk
nltk.download('stopwords')
from nltk.corpus import stopwords
from nltk.tokenize import TreebankWordTokenizer
from tensorflow.keras.preprocessing.text import Tokenizer
from tensorflow.keras.preprocessing.sequence import pad_sequences
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense,LSTM,Embedding
stop_word = set(stopwords.words('english'))
word_tokenizer = TreebankWordTokenizer()
def preprocess(text):
    text = text.lower()
    text = re.sub(r'[^a-z\s]','',text)
    tokens = word_tokenizer.tokenize(text)
    filtered = [word for word in tokens if word not in stop_word]
    return ' '.join(filtered)
preprocess_doc = [preprocess(doc) for doc in titles]
num_tokenizer = Tokenizer(num_words=10000,oov_token='<OOV>')
num_tokenizer.fit_on_texts(preprocess_doc)
seq= num_tokenizer.texts_to_sequences(preprocess_doc)
padded_seq = pad_sequences(seq,maxlen=10,padding='post')
model = Sequential([
    Embedding(input_dim=10000,output_dim=16),
    LSTM(32),
    Dense(1,activation='sigmoid')
])
model.compile(optimizer='adam',loss='binary_crossentropy',metrics=['accuracy'])
model.fit(padded_seq,labels,epochs=50)
def pre_sentiment(user_input):
    user = preprocess(user_input)
    seq_input = num_tokenizer.texts_to_sequences([user])
    padded_input = pad_sequences(seq_input,maxlen=10,padding='post')
    prediction  = model.predict(padded_input).item()
    result = 'CRIMINAL' if prediction>=0.5 else 'NOT CRIMINAL'
    return(f'{result} - Score: {prediction}')
demo = gr.Interface(fn=pre_sentiment,inputs='text',outputs='text',title='Sentiment Analyst')
demo.launch()