Spaces:
Sleeping
Sleeping
import joblib | |
import streamlit as st | |
import json | |
import requests | |
from bs4 import BeautifulSoup | |
# load all the models and vectorizer (global vocabulary) | |
# Seq_model = load_model("LSTM.h5") # Sequential | |
SVM_model = joblib.load("SVM_Linear_Kernel.joblib") # SVM | |
logistic_model = joblib.load("Logistic_Model.joblib") # Logistic | |
vectorizer = joblib.load("vectorizer.joblib") # global vocabulary (used for Logistic, SVC) | |
# tokenizer = joblib.load("tokenizer.joblib") # used for LSTM | |
# Decode label function | |
# {'business': 0, 'entertainment': 1, 'health': 2, 'politics': 3, 'sport': 4} | |
def categorize(input_number): | |
print('receive label encoded', input_number) | |
categories = { | |
0: 'Business', | |
1: 'Entertainment', | |
2: 'Health', | |
3: 'Politics', | |
4: 'Sport' | |
} | |
result = categories.get(input_number) # Ex: Health | |
print('decoded result', result) | |
return result | |
# Web Crawler function | |
def crawURL(url): | |
# Fetch the URL content | |
response = requests.get(url) | |
# Parse the sitemap HTML | |
soup = BeautifulSoup(response.content, 'html.parser') | |
# Find all anchor tags that are children of span tags with class 'sitemap-link' | |
urls = [span.a['href'] for span in soup.find_all('span', class_='sitemap-link') if span.a] | |
# Crawl pages and extract data | |
try: | |
print(f"Crawling page: {url}") | |
# Fetch page content | |
page_response = requests.get(url) | |
page_content = page_response.content | |
# Parse page content with BeautifulSoup | |
soup = BeautifulSoup(page_content, 'html.parser') | |
# Extract data you need from the page | |
author = soup.find("meta", {"name": "author"}).attrs['content'].strip() | |
date_published = soup.find("meta", {"property": "article:published_time"}).attrs['content'].strip() | |
article_section = soup.find("meta", {"name": "meta-section"}).attrs['content'] | |
url = soup.find("meta", {"property": "og:url"}).attrs['content'] | |
headline = soup.find("h1", {"data-editable": "headlineText"}).text.strip() | |
description = soup.find("meta", {"name": "description"}).attrs['content'].strip() | |
keywords = soup.find("meta", {"name": "keywords"}).attrs['content'].strip() | |
text = soup.find(itemprop="articleBody") | |
# Find all <p> tags with class "paragraph inline-placeholder" | |
paragraphs = text.find_all('p', class_="paragraph inline-placeholder") | |
# Initialize an empty list to store the text content of each paragraph | |
paragraph_texts = [] | |
# Iterate over each <p> tag and extract its text content | |
for paragraph in paragraphs: | |
paragraph_texts.append(paragraph.text.strip()) | |
# Join the text content of all paragraphs into a single string | |
full_text = ''.join(paragraph_texts) | |
return full_text | |
except Exception as e: | |
print(f"Failed to crawl page: {url}, Error: {str(e)}") | |
return null | |
# Predict for text category using Models | |
def process_api(text): | |
# Vectorize the text data | |
processed_text = vectorizer.transform([text]) | |
# sequence = tokenizer.texts_to_sequences([text]) | |
# padded_sequence = pad_sequences(sequence, maxlen=1000, padding='post') | |
# Get the predicted result from models | |
Logistic_Predicted = logistic_model.predict(processed_text).tolist() # Logistic Model | |
SVM_Predicted = SVM_model.predict(processed_text).tolist() # SVC Model | |
# Seq_Predicted = Seq_model.predict(padded_sequence) | |
# predicted_label_index = np.argmax(Seq_Predicted) | |
return { | |
'Logistic_Predicted': categorize(int(Logistic_Predicted[0])), | |
'SVM_Predicted': categorize(int(SVM_Predicted[0])), | |
'Article_Content': text | |
} | |
# Using Model to handle and return Category Route | |
def categorize(url): | |
try: | |
article_content = crawURL(url) | |
result = process_api(article_content) | |
return result | |
except Exception as error: | |
if hasattr(error, 'message'): | |
return {"error_message": error.message} | |
else: | |
return {"error_message": error} | |
# Main App | |
url = st.text_input("enter your CNN's URL here") | |
if url: | |
result = categorize(url) | |
article_content = result.get('Article_Content') | |
st.text_area("Article Content", value=article_content, height=400) # render the article content as textarea element | |
st.json(result) |