import time
import nltk
from nltk.sentiment import SentimentIntensityAnalyzer
import gradio as gr
from transformers import AutoTokenizer
from transformers import AutoModelForSequenceClassification
from scipy.special import softmax
import nltk
import matplotlib.pyplot as plt
nltk.download('vader_lexicon')
sia = nltk.sentiment.SentimentIntensityAnalyzer()


def sentiment_analyzer(text, model_name):
  positive_score = 0.0
  negative_score = 0.0
  if model_name == "Vader":
      sentiment_score = sia.polarity_scores(text)
      positive_score = sentiment_score['pos']
      negative_score = sentiment_score['neg']
  elif model_name == "RoBERTa":
      MODEL = f"cardiffnlp/twitter-roberta-base-sentiment"
      tokenizer = AutoTokenizer.from_pretrained(MODEL)
      model = AutoModelForSequenceClassification.from_pretrained(MODEL)
      encoded_text = tokenizer(text, return_tensors='pt')
      output = model(**encoded_text)
      scores = output[0][0].detach().numpy()
      scores = softmax(scores)
      scores_dict = {
          'roberta_neg' : scores[0],
          'roberta_neu' : scores[1],
          'roberta_pos' : scores[2]
      }
      positive_score = float(scores[2])
      negative_score = float(scores[0])

  # print(f"Output positive score: {positive_score}")
  # print(f"Output negative score: {negative_score}")
  return positive_score, negative_score
    
sentiment_analyzer("I am a goog human being","RoBERTa")

input_text = gr.inputs.Textbox(lines=15, label="Input Text")
sentiment_model = gr.inputs.Radio(["Vader", "RoBERTa"], label="Choose Sentiment Analyzer")

outputs = [gr.Slider(minimum=0, maximum=1, step=0.01, label="Positive Sentiment"),
           gr.Slider(minimum=0, maximum=1, step=0.01, label="Negative Sentiment")]

iface = gr.Interface(fn=sentiment_analyzer, inputs=[input_text, sentiment_model], outputs=outputs)

print(time.time())
iface.launch(inline=False)