import gradio as gr
import pandas as pd
import numpy as np

#pip install langdetect
from langdetect import detect

#pip install vaderSentiment
from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer

from sklearn.model_selection import train_test_split
from sklearn.svm import SVR
from sklearn.metrics import mean_squared_error

data = pd.read_csv("modeled_data.csv")
analyzer = SentimentIntensityAnalyzer()

def sample_model(df, regressor, scale=None):
    X = df.drop("rate",axis=1)
    y = df["rate"]
    
    X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.2, random_state=1)
    scaled_X_train, scaled_X_test = X_train, X_test
    if scale != None:
        scaler = scale
        scaled_X_train = pd.DataFrame(scaler.fit_transform(X_train), columns = X_train.columns)
        scaled_X_test = pd.DataFrame(scaler.transform(X_test),columns = X_test.columns)
    
    model = regressor
    model.fit(scaled_X_train, y_train)
    y_pred = model.predict(scaled_X_test)
    
    rmse = np.sqrt(mean_squared_error(y_test, y_pred))
    
    return model, scaled_X_train, scaled_X_test, y_train, y_test 


def user_interaction(comment, model):
    
    negative_score = analyzer.polarity_scores(comment)["neg"]
    neutral_score = analyzer.polarity_scores(comment)["neu"]
    positive_score = analyzer.polarity_scores(comment)["pos"]
    compound_score = analyzer.polarity_scores(comment)["compound"]
    rate_pred = model.predict([[negative_score, neutral_score, positive_score, compound_score]])
    
    return round(negative_score,2), round(neutral_score,2), round(positive_score,2), round(compound_score,2), round(rate_pred[0],2)
    """return (f"\nYour Comment: {comment}\n" + 
           "*"*10 + "Analysis of the Comment" + "*"*10 + "\n" +
           "-"*10 + f"Negativity Score: {negative_score:.2f}" + "-"*10 + "\n" +
           "-"*10 + f"Neutrality Score: {neutral_score:.2f}" + "-"*10 + "\n" +
           "-"*10 + f"Positivity Score: {positive_score:.2f}" + "-"*10 + "\n" +
           "-"*10 + f"Compound Score: {compound_score:.2f}" + "-"*10 + "\n" +
           "*"*43 + "\n"), ("\nThe estimated rating this comment can give" + "\n" + 
           "*"*20 + str(round(rate_pred[0], 2)) + "*"*20 + "\n")"""


def take_input(comment):
    #if (detect(comment) != "en") or (len(comment) < 20):
     #   return "Sorry, your comment does not meet the requirements.\n", "Please check your comment"
    #else:
    cons_tuned_svr, _, _, _, _ = sample_model(data, SVR(C=3, kernel="rbf", tol=0.001))
    return user_interaction(comment, cons_tuned_svr)




with gr.Blocks() as demo:
    gr.Markdown("# AIN311 Project P05 - MOOC Recommendation")
    gr.Markdown("## Generating a Rating from User Comment")
    
    with gr.Column():
        gr.Markdown("""
                    ##### Thanks for your interest and taking your time.
                    ##### Tell us about your personal experience enrolling in this course. Was it the right match for you?
                    ##### (Note: Comment should be written in English and be longer than 20 characters)
                    """)
        input_comment = gr.Textbox(placeholder="Write your comment here...")
        button = gr.Button("What is the Rating I Gave? Click me to Learn")
        gr.Markdown("#### Sentiment Scores of Your Comment")
        negscore = gr.Number(label="Negativity Score")
        neuscore = gr.Number(label="Neutrality Score")
        posscore = gr.Number(label="Positivity Score")
        compscore = gr.Number(label="Compound Score")
        rating = gr.Number(label="Generated Rating from Your Comment")        
    button.click(fn=take_input, inputs=input_comment, outputs=[negscore, neuscore, posscore, compscore, rating])

demo.launch()