Spaces:
Sleeping
Sleeping
File size: 7,660 Bytes
77f904f 69a88fc e024b69 ba5f8d5 84d8e35 f8f4b5f b71cd64 d79611d 75dc5f6 b71cd64 c4db18d f8f4b5f 3c7207a 22da72a af31e8a 3c7207a 192d8ff 3c7207a b71cd64 d6a49a1 b71cd64 9ee5788 d6a49a1 84d8e35 d6a49a1 84d8e35 d6a49a1 84d8e35 d6a49a1 84d8e35 d6a49a1 84d8e35 d6a49a1 84d8e35 d6a49a1 b71cd64 d6a49a1 1548124 b71cd64 d6a49a1 3c7207a b71cd64 d6a49a1 22da72a 52f1084 d6a49a1 22da72a 7e83a81 d6a49a1 a42e0f5 fd5cbde d6a49a1 b71cd64 247496e 4604f63 247496e 4604f63 247496e 192d8ff f8f4b5f 6d50c62 af31e8a 4604f63 22da72a 4604f63 247496e |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 |
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 decodedLabel(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)
# ----------- Debug Logs -----------
logistic_debug = decodedLabel(int(Logistic_Predicted[0]))
svc_debug = decodedLabel(int(SVM_Predicted[0]))
print('Logistic', int(Logistic_Predicted[0]), logistic_debug)
print('SVM', int(SVM_Predicted[0]), svc_debug)
return {
'Logistic_Predicted':decodedLabel(int(Logistic_Predicted[0])),
'SVM_Predicted': decodedLabel(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
st.header('Classification Project')
st.subheader
(
'''
Unsure what category a CNN article belongs to?
Our clever tool can help!
Paste the URL below and press Enter. We'll sort it into one of our 5 categories in a flash! β‘οΈ
'''
)
# Define category information (modify content and bullet points as needed)
categories = {
"Business": [
"Analyze market trends and investment opportunities.",
"Gain insights into company performance and industry news.",
"Stay informed about economic developments and regulations."
],
"Health": [
"Discover healthy recipes and exercise tips.",
"Learn about the latest medical research and advancements.",
"Find resources for managing chronic conditions and improving well-being."
],
"Sport": [
"Follow your favorite sports teams and athletes.",
"Explore news and analysis from various sports categories.",
"Stay updated on upcoming games and competitions."
],
"Politics": [
"Get informed about current political events and policies.",
"Understand different perspectives on political issues.",
"Engage in discussions and debates about politics."
],
"Entertainment": [
"Find recommendations for movies, TV shows, and music.",
"Explore reviews and insights from entertainment critics.",
"Stay updated on celebrity news and cultural trends."
]
}
# Create expanders contain list of category can be classified
for category, content in categories.items():
with st.expander(category, expanded=True):
# Display content as bullet points
for item in content:
st.write(f"- {item}")
# Explain to user why this project is only worked for CNN domain
with st.expander("Tips", expanded=True):
st.write(
'''
This project works best with CNN articles right now.
Our web crawler is like a special tool for CNN's website.
It can't quite understand other websites because they're built differently
'''
)
url = st.text_input("Find your favorite CNN story! Paste the URL here.", placeholder='Ex: https://edition.cnn.com/2012/01/31/health/frank-njenga-mental-health/index.html')
st.divider() # π Draws a horizontal rule
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.divider() # π Draws a horizontal rule
st.json({
"Logistic": result.get("Logistic_Predicted"),
"SVC": result.get("SVM_Predicted")
})
st.divider() # π Draws a horizontal rule
# Category labels and corresponding counts
categories = ["Sport", "Health", "Entertainment", "Politics", "Business"]
counts = [5638, 4547, 2658, 2461, 1362]
# Optional: Add a chart title
st.title("Training Data Category Distribution")
# Optional: Display additional information
st.write("Here's a breakdown of the number of articles in each category:")
for category, count in zip(categories, counts):
st.write(f"- {category}: {count}")
# Create the bar chart
st.bar_chart(data=dict(zip(categories, counts))) |