from transformers import *
import gradio as gr
import nltk
from nltk.tokenize import sent_tokenize
from language_tool_python import LanguageTool
from nltk.sentiment import SentimentIntensityAnalyzer
import torch

nltk.download('vader_lexicon')
nltk.download('stopwords')

model = PegasusForConditionalGeneration.from_pretrained("tuner007/pegasus_paraphrase")
tokenizer = PegasusTokenizerFast.from_pretrained("tuner007/pegasus_paraphrase")
def get_paraphrased_sentences(paragraph):
    num_return_sequences = 1
    num_beams = 1
    # Split the paragraph into sentences
    sentences = paragraph.split(". ")
    
    paraphrased_sentences = []
    for sentence in sentences:
        # Tokenize the sentence
        inputs = tokenizer([sentence], truncation=True, padding="longest", return_tensors="pt")
        # Generate the paraphrased sentences
        outputs = model.generate(
            **inputs,
            num_beams=num_beams,
            num_return_sequences=num_return_sequences,
        )
        # Decode the generated sentences using the tokenizer to get them back to text
        paraphrased = tokenizer.batch_decode(outputs, skip_special_tokens=True)
        paraphrased_sentences.extend(paraphrased)
    
    return paraphrased_sentences
# Initialize LanguageTool and SentimentIntensityAnalyzer objects
tool = LanguageTool('en-US')
sia = SentimentIntensityAnalyzer()
def grammar_check(text):
    matches = tool.check(text)
    corrected_text = tool.correct(text)
    return corrected_text, matches

def analyze_sentiment(text):
    sentiment_scores = sia.polarity_scores(text)
    sentiment = ""
    if sentiment_scores['compound'] > 0:
        sentiment = "Positive"
    elif sentiment_scores['compound'] == 0:
        sentiment = "Neutral"
    else:
        sentiment = "Negative"
    return sentiment

def extract_wrong_words(matches):
    wrong_words = set()
    for match in matches:
        wrong_words.update(match.replacements)
    return wrong_words

def dropdown_select(text, output_type):
    corrected_text, matches = grammar_check(text)
    sentiment_result = analyze_sentiment(corrected_text)
    words = corrected_text.split()
    wrong_words = extract_wrong_words(matches)
    if output_type == "Grammar":
        return corrected_text, None, sentiment_result, len(words), wrong_words
    elif output_type == "Paraphrase":
        paraphrase_text = get_paraphrased_sentences(corrected_text)
        return None, paraphrase_text, sentiment_result, len(words), wrong_words
    elif output_type == "All":
        paraphrase_text = get_paraphrased_sentences(corrected_text)
        return corrected_text, paraphrase_text, sentiment_result, len(words), wrong_words
    else:
        return None,None,None,None,None,

import gradio as gr
iface = gr.Interface(
    fn=dropdown_select,
    inputs=[
        gr.inputs.Textbox(placeholder="Enter your text here..."),
        gr.inputs.Dropdown(["All", "Grammar", "Paraphrase"], label="Select Output Type", default=None)
    ],
   outputs=[
        gr.outputs.Textbox(label="Modified Grammar"),
        gr.outputs.Textbox(label="Paraphrased Content"),
        gr.outputs.Textbox(label="Sentiment Analysis"),
        gr.outputs.Textbox(label="Total Words Count"),
        gr.outputs.Textbox(label="Detected Wrong Words")
    ],
    title="CSharpGrammarly",
    description="CSHARPGRAMMARLYTOOL",
    allow_flagging=False  # Disable flagging for simplicity
)

iface.launch(debug=True)