Web Scraping with Langchain and html2text

#python #programming #beginners #tutorial

Introduction

In this blog post, you will be guided on how to perform a simple web scrapping using the available open source python packages. We are going to make use of the langchain and html2text.

Hands-on

First, we need to make sure to install the langchain and html2text packages.

!pip install -q langchain playwright beautifulsoup4 html2text

Here's the code snippet for accomplishing the web scrapping. The following code is utilizing the langchain's AsyncHtmlLoader and the Html2TextTransformer from html2text package for the extraction of HTML to Text.

import html2text
from langchain.document_loaders import AsyncHtmlLoader
from langchain.document_transformers import Html2TextTransformer

async def do_webscraping(link):
    try:
        urls = [link]
        loader = AsyncHtmlLoader(urls)
        docs = loader.load()

        html2text_transformer = Html2TextTransformer()
        docs_transformed = html2text_transformer.transform_documents(docs)

        if docs_transformed != None and len(docs_transformed) > 0:
            metadata = docs_transformed[0].metadata
            title = metadata.get('title', '')
            return {
                'summary': docs_transformed[0].page_content,
                'title': title,
                'metadata': metadata,
                'clean_content': html2text.html2text(docs_transformed[0].page_content)
            }
        else:
            return None

    except Exception as e:
        print(f"An unexpected error occurred: {e}")
        return None

Let's see the real code usages now. Assume below are the URLs for which you wish to do the web scrapping.

google_search_results = ['https://www.yelp.com/search?cflt=seafood&find_loc=Mountain+View%2C+CA+94043',
 'https://www.yelp.com/search?cflt=seafood&find_loc=Mountain+View%2C+CA',
 'https://www.opentable.com/cuisine/best-seafood-restaurants-mountain-view-ca']

Here's the logic for interacting and performing the web scrapping.

for link in google_search_results:
  print(link)
  response = await do_webscraping(link)
  if response != None:
    structured_response.append(response)