Unstructured vs LlamaParse comparison

from unstructured.partition.pdf import partition_pdf
import os

# Path to PDF file
pdf_path = "sample.pdf"

# Extract elements from PDF
elements = partition_pdf(filename=pdf_path)

# Simple extraction of title and author from metadata or first elements
title = None
author = None
for el in elements:
    if el.metadata and 'Title' in el.metadata:
        title = el.metadata['Title']
    if el.metadata and 'Author' in el.metadata:
        author = el.metadata['Author']
    if title and author:
        break

print(f"Title: {title}")
print(f"Author: {author}")

import os
from llamaparse import LlamaParse
from openai import OpenAI

# Initialize OpenAI client
client = OpenAI(api_key=os.environ["OPENAI_API_KEY"])

# Initialize LlamaParse with the LLM client
parser = LlamaParse(llm_client=client)

# Load PDF text (assume text extracted separately)
pdf_text = """
Example Document Title\nAuthor: Jane Doe\nThis is the content of the document...
"""

# Define prompt for semantic extraction
prompt = f"Extract the title and author from the following document text:\n{pdf_text}\nReturn as JSON with keys 'title' and 'author'."

# Parse document
result = parser.parse(prompt)

print(result)  # Expected: {'title': 'Example Document Title', 'author': 'Jane Doe'}