By Aaron LeBlanc, Founder & CEO, Hypelocal

➡️ https://zapier.com/experts/hypelocal

Presentation Deck

[https://docs.google.com/presentation/d/1Je5Wg75GvSvFc22X51YhPhK33_3jOFQRphwTZsDCi4Q/edit?usp=sharing](https://docs.google.com/presentation/d/1Je5Wg75GvSvFc22X51YhPhK33_3jOFQRphwTZsDCi4Q/preview?usp=sharing)

What is Unstructured Data?

Use Case 1: Structuring raw unstructured website content

• Process:

  

• Code Step to Scrape Raw Website Content from URL

  import requests
  import re
  import logging
  
  # Configure logging
  logging.basicConfig(level=logging.INFO)
  
  def extract_text_from_html(html_content):
      """
      Extracts and cleans text from HTML content.
      """
      try:
          # Remove script and style elements
          html_content = re.sub(r'<(script|style).*?>.*?</\\1>', '', html_content, flags=re.DOTALL)
          # Remove HTML tags
          text = re.sub(r'<.*?>', ' ', html_content)
          # Collapse whitespace
          text = re.sub(r'\\s+', ' ', text)
          return text.strip()
      except Exception as e:
          logging.error(f"Failed to extract text from HTML: {e}")
          return ""
  
  def scrape_website_content(website_url):
      """
      Scrapes raw content from the given website URL.
      """
      try:
          # Ensure the URL is prefixed with http:// or https://
          if not website_url.startswith(('http://', 'https://')):
              website_url = 'http://' + website_url
          response = requests.get(website_url)
          response.raise_for_status()
          raw_text = extract_text_from_html(response.content.decode('utf-8'))
          return raw_text
      except requests.RequestException as e:
          logging.error(f"Failed to fetch URL {website_url}: {e}")
          return ""
  
  # Get the website URL from input data
  website_url = input_data.get('websiteUrl')
  
  if website_url:
      raw_content = scrape_website_content(website_url)
      if raw_content:
          output = {'raw_content': raw_content}
      else:
          output = {'error': 'Failed to extract content from the website.'}
  else:
      output = {'error': 'No website URL provided.'}
  
  return output
  
  

• ChatGPT Prompt Template for Structuring Raw Unstructured Website Content

  Review the raw content here:
  ----
  Website url: {{249095056__fields__websiteUrl}}. # From Zapier Chrome Push Trigger Step
  Content: {{249095057__raw_content}}. # From Zapier Chrome Push Trigger Step
  ----
  
  Extract and structure the data based on the following schema:
  ---
  - 'businessName': Extracted from the business name.
  - 'businessStreet': Extracted from business address.
  - 'businessCity': Extracted from business address.
  - 'businessState': Extracted from business address (state or province).
  - 'businessCountry': Extracted from business address.
  - 'businessPostalCode': Extracted from business address.
  - 'businessPhone': Extracted from the business phone number.
  - 'businessEmail': Extracted from the business email address.
  - 'websiteUrl': Extracted from the Company website URL.
  - 'businessDescription': A detailed and specific description of the business and what they do. Three sentences or less.
  - 'businessIndustryName': The business industry (Name only, no code) using NAICS standards.
  ---
  
  - Once the data is gathered, check your response for accuracy against the provided instructions.
  - Unless otherwise told in the schema, all fields are string fields.
  - This data will be used to go into a CRM via automation. The data needs to needs to be accurate, clean, and contextual.
  
  ++++
  Output response in JSON code format with no leading characters.  Your reply will be used as a JSON payload. Don't include (```json```).
  ++++
  

  • ChatGPT Settings:

    • Model = gpt-4o

    • Memory Key = blank

    • Image = blank

    • User Name = Expert Sales Person

    • Assistant Name = Expert Sales Person

    • Assistant Instructions =

      You are a helpful Sales assistant that specializes in parsing unstructured content into a structured format.
      

    • Max Tokens = 1024

    • Temperature = 0.5

    • Top P = 1

• Parse JSON Payload Code Step

  • Input Data - payLoad = ChatGPT Response

  var obj = JSON.parse(inputData.payLoad);
  return obj;
  

• Setup Action steps as needed from there.

Use Case 2: Structuring raw unstructured email content

• Process:

  

• ChatGPT Prompt Template for Structuring Unstructured Email Content

  Review the raw content here:
  ----
  Subject: {{249105134__raw__Subject}}. # From Zapier Email Trigger Step
  Body plain: {{249105134__body_plain}}. # From Zapier Email Trigger Step
  ----
  
  Extract and structure the data based on the following schema:
  ---
  - 'businessName': Extracted from the business name.
  - 'businessStreet': Extracted from business address.
  - 'businessCity': Extracted from business address.
  - 'businessState': Extracted from business address (state or province).
  - 'businessCountry': Extracted from business address.
  - 'businessPostalCode': Extracted from business address.
  - 'businessPhone': Extracted from the business phone number.
  - 'businessEmail': Extracted from the business email address.
  - 'websiteUrl': Extracted from the URL in the prospects email.
  - 'businessDescription': A detailed and specific description of the business and what they do. Three sentences or less.
  - 'businessIndustryName': Infer the business industry using NAICS standards.
  - 'contactFirstName': Extracted from the contact's name.
  - 'contactLastName': Extracted from the contact's name.
  - 'contactCellPhone': Extracted from the contact's cell phone number.
  - 'contactEmail': Extracted from the contact's email address.
  - 'contactTitle': Extracted from the contact's title or position.
  - 'budget': Extracted budget information related to the business or project.
  - 'prospectInterest': Extracted interest level or area of interest of the prospect.
  - 'leadDescription': A brief description of the lead, summarizing the potential opportunity or interest.
  - 'whenToContact': Extracted preferred time or date to contact the lead.
  ---
  
  - Once the data is gathered, check your response for accuracy against the provided instructions.
  - Unless otherwise told in the schema, all fields are string fields.
  - This data will be used to go into a CRM via automation. The data needs to needs to be accurate, clean, and contextual.
  
  ++++
  Output response in JSON code format with no leading characters.  Your reply will be used as a JSON payload. Don't include (```json```).
  ++++
  

  • ChatGPT Settings:

    • Model = gpt-4

    • Memory Key = blank

    • Image = blank

    • User Name = Company Admin

    • Assistant Name = Company Admin Assistant and Email Parser

    • Assistant Instructions =

      You are a helpful Company assistant that specializes in parsing unstructured content into a structured format.
      

    • Max Tokens = 1024

    • Temperature = 0.5

    • Top P = 1

• Parse JSON Payload Code Step

  • Input Data - payLoad = ChatGPT Response

  var obj = JSON.parse(inputData.payLoad);
  return obj;
  

• Setup Action steps as needed from there.