app.py

import streamlit as st
from selenium import webdriver
from selenium.webdriver.common.by import By
from selenium.webdriver.chrome.options import Options
from selenium.webdriver.chrome.service import Service
import pandas as pd
from selenium.webdriver.support.ui import WebDriverWait
from selenium.webdriver.support import expected_conditions as EC
import time
from webdriver_manager.chrome import ChromeDriverManager
from webdriver_manager.chrome import ChromeType
import transformers
import torch
import plotly.express as px

st.subheader("YouTube Comments Sentiment Analysis", divider="red")
tokenizer = transformers.DistilBertTokenizer.from_pretrained("tabularisai/robust-sentiment-analysis")
model = transformers.DistilBertForSequenceClassification.from_pretrained("tabularisai/robust-sentiment-analysis")

if 'url_count' not in st.session_state:
    st.session_state['url_count'] = 0

max_attempts = 2

def update_url_count():
    st.session_state['url_count'] += 1

def clear_question():
    st.session_state["url"] = ""

url = st.text_input("Enter YouTube URL:", key="url")
st.button("Clear question", on_click=clear_question)

if st.button("Sentiment Analysis", type="secondary"):
    if st.session_state['url_count'] < max_attempts:
        if url:
            with st.spinner("Wait for it...", show_time=True):
                options = Options()
                options.add_argument("--headless")
                options.add_argument("--disable-gpu")
                options.add_argument("--no-sandbox")
                options.add_argument("--disable-dev-shm-usage")
                options.add_argument("--start-maximized")
                service = Service(ChromeDriverManager(chrome_type=ChromeType.CHROMIUM).install())
                driver = webdriver.Chrome(service=service, options=options)
                data = []
                wait = WebDriverWait(driver, 30)
                driver.get(url)

                placeholder = st.empty()
                progress_bar = st.progress(0)

                for item in range(30):
                    try:
                        driver.execute_script("window.scrollBy(0, 500);")
                        wait.until(EC.presence_of_element_located((By.CSS_SELECTOR, "#content #content-text")))
                        placeholder.text(f"Scrolled {item + 1} times")
                        progress_bar.progress((item + 1) / 30)
                        time.sleep(1) #Increased wait time for dynamic loading
                    except Exception as e:
                        st.error(f"Exception during scrolling: {e}")
                        break

                placeholder.text("Scrolling complete.")
                progress_bar.empty()

                try:
                    wait.until(EC.presence_of_element_located((By.CSS_SELECTOR, "#contents #contents")))
                    comments = driver.find_elements(By.CSS_SELECTOR, "#content #content-text")
                    for comment in comments:
                        timestamp = None
                        try:
                            comment_text = comment.text
                            date_match = re.search(r'\d+ (day|week|month|year)s? ago', comment_text) #Example regex.
                            if date_match:
                                timestamp = date_match.group(0)
                        except Exception as e:
                            st.error(f"Error extracting date with regex: {e}")
                        data.append({"Comment": comment.text, "comment_date": timestamp})


                except Exception as e:
                    st.error(f"Exception during comment extraction: {e}")
                driver.quit()
                df = pd.DataFrame(data, columns=["Comment", "comment_date"])

                if not df.empty and not df['Comment'].tolist() == []:
                    st.dataframe(df)
                    inputs = tokenizer(df['Comment'].tolist(), return_tensors="pt", padding=True, truncation=True)
                    with torch.no_grad():
                        logits = model(**inputs).logits
                        predicted_probabilities = torch.nn.functional.softmax(logits, dim=-1)
                        predicted_labels = predicted_probabilities.argmax(dim=1)
                        results = []
                        for i, label in enumerate(predicted_labels):
                            results.append({'Review Number': i + 1, 'Sentiment': model.config.id2label[label.item()]})
                        sentiment_df = pd.DataFrame(results)

                    value_counts1 = sentiment_df['Sentiment'].value_counts().rename_axis('Sentiment').reset_index(name='count')
                    final_df = value_counts1
                    tab1, tab2 = st.tabs(["Pie Chart", "Bar Chart"])
                    with tab1:
                        fig1 = px.pie(final_df, values='count', names='Sentiment', hover_data=['count'], labels={'count': 'count'})
                        fig1.update_traces(textposition='inside', textinfo='percent+label')
                        st.plotly_chart(fig1)

                    result = pd.concat([df, sentiment_df], axis=1)
                    st.dataframe(result)

                    with tab2:
                        fig2 = px.bar(result, x="Sentiment", y="comment_date", color="Sentiment")
                        st.plotly_chart(fig2)

                    csv = result.to_csv(index=False)
                    st.download_button(label="Download data as CSV", data=csv, file_name='Summary of the results.csv', mime='text/csv')
                else:
                    st.warning("No comments were scraped. Sentiment analysis could not be performed.")

        else:
            st.warning("Please enter a URL.")
    else:
        st.warning(f"You have reached the maximum URL attempts ({max_attempts}).")

if 'url_count' in st.session_state:
    st.write(f"URL pasted {st.session_state['url_count']} times.")