CourseSearch / preprocess.py
Pontonkid's picture
Create preprocess.py
4e214e0 verified
import time
import numpy as np
import pandas as pd
import requests
import streamlit as st
from bs4 import BeautifulSoup
from sentence_transformers import SentenceTransformer, util
def preprocess():
# Base URL for navigation
base_url = 'https://courses.analyticsvidhya.com/collections/courses?page='
course_list_url = "https://courses.analyticsvidhya.com/"
# List to hold course data
courses = []
page_number = 1 # Start with the first page
while True:
# Construct URL for the current page
current_page_url = base_url + str(page_number)
print(f"Processing page {page_number}...")
# Get the current page content
response = requests.get(current_page_url)
if response.status_code != 200:
print(f"Failed to fetch page {page_number}. Status code: {response.status_code}")
break
soup = BeautifulSoup(response.content, 'html.parser')
# Find all course cards
course_cards = soup.find_all('li', class_='products__list-item')
if not course_cards:
print("No more courses found. Ending extraction.")
break
# Extract course data from each card
for course_card in course_cards:
title_tag = course_card.find('h3')
link_tag = course_card.find('a')
if title_tag and link_tag: # Check if both title and link exist
title = title_tag.text.strip()
course_link = link_tag['href']
# Construct full course URL (assume relative links)
course_url = course_list_url.rstrip('/') + course_link
# Visit each course link to get the description
course_response = requests.get(course_url)
if course_response.status_code == 200:
course_soup = BeautifulSoup(course_response.content, 'html.parser')
description_tag = course_soup.find('div', class_='fr-view') # Adjust based on actual class or tag
description = description_tag.text.strip() if description_tag else 'No description available'
curriculum_tag = course_soup.find('ul', class_='course-curriculum__chapter-content') # Adjust based on actual class or tag
curriculum = curriculum_tag.text.strip() if curriculum_tag else 'No curriculum available'
#enroll_tag = course_soup.find('article', class_='section__content section__content___ae733') # Adjust based on actual class or tag
#enroll = enroll_tag.text.strip() if enroll_tag else 'No enroll available'
instructor_tag = course_soup.find('section', class_='text-image section-height__medium section__content-alignment--left text-image___07200') # Adjust based on actual class or tag
instructor = instructor_tag.text.strip() if instructor_tag else 'No instructor available'
# Append the data to the list
courses.append({'title': title, 'description': description, 'Course curriculum': curriculum, 'About the Instructor': instructor})
else:
print(f"Failed to fetch course page: {course_url}")
# Sleep to avoid overwhelming the server (optional)
time.sleep(1)
else:
print("Skipped a course card due to missing title or link.")
# Move to the next page
page_number += 1
# break
# Save the collected data to a CSV file
df = pd.DataFrame(courses)
df.to_csv('courses.csv', index=False)
print("Data collection complete. Saved to courses.csv.")
# Load the data
df = pd.read_csv('courses.csv')
# Combine relevant text fields for embedding (e.g., title, description, curriculum)
df['combined_text'] = df['title'] + ' ' + df['description'] + ' ' + df['Course curriculum'] + ' ' + df['About the Instructor']
# Load a pre-trained model for embeddings
model = SentenceTransformer('all-MiniLM-L6-v2')
# Create embeddings for each course
embeddings = model.encode(df['combined_text'].tolist(), convert_to_tensor=True)
# Save embeddings and DataFrame for later use
np.save('course_embeddings.npy', embeddings)
df.to_csv('courses_with_embeddings.csv', index=False)
# Load embeddings and DataFrame
embeddings = np.load('course_embeddings.npy')
df = pd.read_csv('courses_with_embeddings.csv')