import streamlit as st import streamlit.components.v1 as com #import libraries from transformers import AutoModelForSequenceClassification,AutoTokenizer, AutoConfig import numpy as np #convert logits to probabilities from scipy.special import softmax #import the model tokenizer = AutoTokenizer.from_pretrained('bert-base-cased') model_path = f"UholoDala/tweet_sentiments_analysis" config = AutoConfig.from_pretrained(model_path) model = AutoModelForSequenceClassification.from_pretrained(model_path) #Set the page configs st.set_page_config(page_title='Sentiments Analysis',page_icon='😎',layout='wide') #welcome Animation com.iframe("https://embed.lottiefiles.com/animation/149093") st.markdown('

Tweet Sentiments

',unsafe_allow_html=True) #Create a form to take user inputs with st.form(key='tweet',clear_on_submit=True): text=st.text_area('Copy and paste a tweet or type one',placeholder='I find it quite amusing how people ignore the effects of not taking the vaccine') submit=st.form_submit_button('submit') #create columns to show outputs col1,col2,col3=st.columns(3) col1.title('Sentiment Emoji') col2.title('How this user feels about the vaccine') col3.title('Confidence of this prediction') if submit: print('submitted') #pass text to preprocessor def preprocess(text): #initiate an empty list new_text = [] #split text by space for t in text.split(" "): #set username to @user t = '@user' if t.startswith('@') and len(t) > 1 else t #set tweet source to http t = 'http' if t.startswith('http') else t #store text in the list new_text.append(t) #change text from list back to string return " ".join(new_text) #pass text to model #change label id config.id2label = {0: 'NEGATIVE', 1: 'NEUTRAL', 2: 'POSITIVE'} text = preprocess(text) # PyTorch-based models encoded_input = tokenizer(text, return_tensors='pt') output = model(**encoded_input) scores = output[0][0].detach().numpy() scores = softmax(scores) #Process scores ranking = np.argsort(scores) ranking = ranking[::-1] l = config.id2label[ranking[0]] s = scores[ranking[0]] #output if l=='NEGATIVE': with col1: com.iframe("https://embed.lottiefiles.com/animation/125694") col2.write('Negative') col3.write(f'{s}%') elif l=='POSITIVE': with col1: com.iframe("https://embed.lottiefiles.com/animation/148485") col2.write('Positive') col3.write(f'{s:.2%}') else: with col1: com.iframe("https://embed.lottiefiles.com/animation/136052") col2.write('Neutral') col3.write(f'{s}%')